Datasets:

---

ftopal

/

huggingface-datasets-processed

like 0

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:crowdsourced", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:1K<n<10K", "source_datasets:original", "language:fr", "license:cc-by-4.0", "arxiv:2206.09917", "region:us" ]

[ "2206.09917" ]

[ "fr" ]

[ "# Dataset Card for Multilingual HateCheck", "## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL", "## Dataset Structure\n\nThe csv format mostly matches the original HateCheck data, with some adjustments for specific languages.\n\nmhc_case_id\nThe test case ID that is unique to each test case across languages (e.g., \"mandarin-1305\")\n\nfunctionality\nThe shorthand for the functionality tested by the test case (e.g, \"target_obj_nh\"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.\n\ntest_case\nThe test case text.\n\nlabel_gold\nThe gold standard label (\"hateful\" or \"non-hateful\") of the test case. All test cases within a given functionality have the same gold standard label.\n\ntarget_ident\nWhere applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.\n\nref_case_id\nFor hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.\n\nref_templ_id\nThe equivalent to ref_case_id, but for template IDs.\n\ntempl_id\nThe ID of the template from which the test case was generated.\n\ncase_templ\nThe template from which the test case was generated (where applicable).\n\ngender_male and gender_female\nFor gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.\n\nlabel_annotated\nA list of labels given by the three annotators who reviewed the test case (e.g., \"['hateful', 'hateful', 'hateful']\").\n\nlabel_annotated_maj\nThe majority vote of the three annotators (e.g., \"hateful\"). In some cases this differs from the gold label given by our language experts.\n\ndisagreement_in_case\nTrue if label_annotated_maj does not match label_gold for the entry.\n\ndisagreement_in_template\nTrue if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC." ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-crowdsourced #language_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-French #license-cc-by-4.0 #arxiv-2206.09917 #region-us \n", "# Dataset Card for Multilingual HateCheck", "## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL", "## Dataset Structure\n\nThe csv format mostly matches the original HateCheck data, with some adjustments for specific languages.\n\nmhc_case_id\nThe test case ID that is unique to each test case across languages (e.g., \"mandarin-1305\")\n\nfunctionality\nThe shorthand for the functionality tested by the test case (e.g, \"target_obj_nh\"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.\n\ntest_case\nThe test case text.\n\nlabel_gold\nThe gold standard label (\"hateful\" or \"non-hateful\") of the test case. All test cases within a given functionality have the same gold standard label.\n\ntarget_ident\nWhere applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.\n\nref_case_id\nFor hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.\n\nref_templ_id\nThe equivalent to ref_case_id, but for template IDs.\n\ntempl_id\nThe ID of the template from which the test case was generated.\n\ncase_templ\nThe template from which the test case was generated (where applicable).\n\ngender_male and gender_female\nFor gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.\n\nlabel_annotated\nA list of labels given by the three annotators who reviewed the test case (e.g., \"['hateful', 'hateful', 'hateful']\").\n\nlabel_annotated_maj\nThe majority vote of the three annotators (e.g., \"hateful\"). In some cases this differs from the gold label given by our language experts.\n\ndisagreement_in_case\nTrue if label_annotated_maj does not match label_gold for the entry.\n\ndisagreement_in_template\nTrue if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC." ]

[ 107, 11, 191, 568 ]

[ "passage: TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-crowdsourced #language_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-French #license-cc-by-4.0 #arxiv-2206.09917 #region-us \n# Dataset Card for Multilingual HateCheck## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL" ]

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:crowdsourced", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:1K<n<10K", "source_datasets:original", "language:nl", "license:cc-by-4.0", "arxiv:2206.09917", "region:us" ]

[ "2206.09917" ]

[ "nl" ]

[ "# Dataset Card for Multilingual HateCheck", "## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL", "## Dataset Structure\n\nThe csv format mostly matches the original HateCheck data, with some adjustments for specific languages.\n\nmhc_case_id\nThe test case ID that is unique to each test case across languages (e.g., \"mandarin-1305\")\n\nfunctionality\nThe shorthand for the functionality tested by the test case (e.g, \"target_obj_nh\"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.\n\ntest_case\nThe test case text.\n\nlabel_gold\nThe gold standard label (\"hateful\" or \"non-hateful\") of the test case. All test cases within a given functionality have the same gold standard label.\n\ntarget_ident\nWhere applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.\n\nref_case_id\nFor hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.\n\nref_templ_id\nThe equivalent to ref_case_id, but for template IDs.\n\ntempl_id\nThe ID of the template from which the test case was generated.\n\ncase_templ\nThe template from which the test case was generated (where applicable).\n\ngender_male and gender_female\nFor gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.\n\nlabel_annotated\nA list of labels given by the three annotators who reviewed the test case (e.g., \"['hateful', 'hateful', 'hateful']\").\n\nlabel_annotated_maj\nThe majority vote of the three annotators (e.g., \"hateful\"). In some cases this differs from the gold label given by our language experts.\n\ndisagreement_in_case\nTrue if label_annotated_maj does not match label_gold for the entry.\n\ndisagreement_in_template\nTrue if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC." ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-crowdsourced #language_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-Dutch #license-cc-by-4.0 #arxiv-2206.09917 #region-us \n", "# Dataset Card for Multilingual HateCheck", "## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL", "## Dataset Structure\n\nThe csv format mostly matches the original HateCheck data, with some adjustments for specific languages.\n\nmhc_case_id\nThe test case ID that is unique to each test case across languages (e.g., \"mandarin-1305\")\n\nfunctionality\nThe shorthand for the functionality tested by the test case (e.g, \"target_obj_nh\"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.\n\ntest_case\nThe test case text.\n\nlabel_gold\nThe gold standard label (\"hateful\" or \"non-hateful\") of the test case. All test cases within a given functionality have the same gold standard label.\n\ntarget_ident\nWhere applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.\n\nref_case_id\nFor hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.\n\nref_templ_id\nThe equivalent to ref_case_id, but for template IDs.\n\ntempl_id\nThe ID of the template from which the test case was generated.\n\ncase_templ\nThe template from which the test case was generated (where applicable).\n\ngender_male and gender_female\nFor gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.\n\nlabel_annotated\nA list of labels given by the three annotators who reviewed the test case (e.g., \"['hateful', 'hateful', 'hateful']\").\n\nlabel_annotated_maj\nThe majority vote of the three annotators (e.g., \"hateful\"). In some cases this differs from the gold label given by our language experts.\n\ndisagreement_in_case\nTrue if label_annotated_maj does not match label_gold for the entry.\n\ndisagreement_in_template\nTrue if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC." ]

[ 107, 11, 191, 568 ]

[ "passage: TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-crowdsourced #language_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-Dutch #license-cc-by-4.0 #arxiv-2206.09917 #region-us \n# Dataset Card for Multilingual HateCheck## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL" ]

[ "task_categories:text-classification", "task_ids:hate-speech-detection", "annotations_creators:crowdsourced", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:1K<n<10K", "source_datasets:original", "language:ar", "license:cc-by-4.0", "arxiv:2206.09917", "region:us" ]

[ "2206.09917" ]

[ "ar" ]

[ "# Dataset Card for Multilingual HateCheck", "## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL", "## Dataset Structure\n\nThe csv format mostly matches the original HateCheck data, with some adjustments for specific languages.\n\nmhc_case_id\nThe test case ID that is unique to each test case across languages (e.g., \"mandarin-1305\")\n\nfunctionality\nThe shorthand for the functionality tested by the test case (e.g, \"target_obj_nh\"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.\n\ntest_case\nThe test case text.\n\nlabel_gold\nThe gold standard label (\"hateful\" or \"non-hateful\") of the test case. All test cases within a given functionality have the same gold standard label.\n\ntarget_ident\nWhere applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.\n\nref_case_id\nFor hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.\n\nref_templ_id\nThe equivalent to ref_case_id, but for template IDs.\n\ntempl_id\nThe ID of the template from which the test case was generated.\n\ncase_templ\nThe template from which the test case was generated (where applicable).\n\ngender_male and gender_female\nFor gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.\n\nlabel_annotated\nA list of labels given by the three annotators who reviewed the test case (e.g., \"['hateful', 'hateful', 'hateful']\").\n\nlabel_annotated_maj\nThe majority vote of the three annotators (e.g., \"hateful\"). In some cases this differs from the gold label given by our language experts.\n\ndisagreement_in_case\nTrue if label_annotated_maj does not match label_gold for the entry.\n\ndisagreement_in_template\nTrue if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC." ]

[ "TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-crowdsourced #language_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-Arabic #license-cc-by-4.0 #arxiv-2206.09917 #region-us \n", "# Dataset Card for Multilingual HateCheck", "## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL", "## Dataset Structure\n\nThe csv format mostly matches the original HateCheck data, with some adjustments for specific languages.\n\nmhc_case_id\nThe test case ID that is unique to each test case across languages (e.g., \"mandarin-1305\")\n\nfunctionality\nThe shorthand for the functionality tested by the test case (e.g, \"target_obj_nh\"). The same functionalities are tested in all languages, except for Mandarin and Arabic, where non-Latin script required adapting the tests for spelling variations.\n\ntest_case\nThe test case text.\n\nlabel_gold\nThe gold standard label (\"hateful\" or \"non-hateful\") of the test case. All test cases within a given functionality have the same gold standard label.\n\ntarget_ident\nWhere applicable, the protected group that is targeted or referenced in the test case. All HateChecks cover seven target groups, but their composition varies across languages.\n\nref_case_id\nFor hateful cases, where applicable, the ID of the hateful case which was perturbed to generate this test case. For non-hateful cases, where applicable, the ID of the hateful case which is contrasted by this test case.\n\nref_templ_id\nThe equivalent to ref_case_id, but for template IDs.\n\ntempl_id\nThe ID of the template from which the test case was generated.\n\ncase_templ\nThe template from which the test case was generated (where applicable).\n\ngender_male and gender_female\nFor gender-inflected languages (French, Spanish, Portuguese, Hindi, Arabic, Italian, Polish, German), only for cases where gender inflection is relevant, separate entries for gender_male and gender_female replace case_templ.\n\nlabel_annotated\nA list of labels given by the three annotators who reviewed the test case (e.g., \"['hateful', 'hateful', 'hateful']\").\n\nlabel_annotated_maj\nThe majority vote of the three annotators (e.g., \"hateful\"). In some cases this differs from the gold label given by our language experts.\n\ndisagreement_in_case\nTrue if label_annotated_maj does not match label_gold for the entry.\n\ndisagreement_in_template\nTrue if the test case is generated from an IDENT template and there is at least one case with disagreement_in_case generated from the same template. This can be used to exclude entire templates from MHC." ]

[ 106, 11, 191, 568 ]

[ "passage: TAGS\n#task_categories-text-classification #task_ids-hate-speech-detection #annotations_creators-crowdsourced #language_creators-expert-generated #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-Arabic #license-cc-by-4.0 #arxiv-2206.09917 #region-us \n# Dataset Card for Multilingual HateCheck## Dataset Description\n\nMultilingual HateCheck (MHC) is a suite of functional tests for hate speech detection models in 10 different languages: Arabic, Dutch, French, German, Hindi, Italian, Mandarin, Polish, Portuguese and Spanish.\nFor each language, there are 25+ functional tests that correspond to distinct types of hate and challenging non-hate.\nThis allows for targeted diagnostic insights into model performance.\n\nFor more details, please refer to our paper about MHC, published at the 2022 Workshop on Online Abuse and Harms (WOAH) at NAACL 2022. If you are using MHC, please cite our work!\n- Paper: Röttger et al. (2022) - Multilingual HateCheck: Functional Tests for Multilingual Hate Speech Detection Models. URL\n- Repository: URL\n- Point of Contact: paul@URL" ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: tals/albert-base-vitaminc\n* Dataset: tals/vitaminc\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @SalimHF for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: tals/albert-base-vitaminc\n* Dataset: tals/vitaminc\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @SalimHF for evaluating this model." ]

[ 13, 82, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: tals/albert-base-vitaminc\n* Dataset: tals/vitaminc\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @SalimHF for evaluating this model." ]

[ "region:us" ]

[]

[]

[ "# Dataset Card for FCD", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nFCD dataset", "### Supported Tasks and Leaderboards\n\nNLP", "### Languages\n\nen", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ "TAGS\n#region-us \n", "# Dataset Card for FCD", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:", "### Dataset Summary\n\nFCD dataset", "### Supported Tasks and Leaderboards\n\nNLP", "### Languages\n\nen", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information" ]

[ 6, 7, 112, 24, 10, 12, 5, 6, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6 ]

[ "passage: TAGS\n#region-us \n# Dataset Card for FCD## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information## Dataset Description\n\n- Homepage: \n- Repository: \n- Paper: \n- Leaderboard: \n- Point of Contact:### Dataset Summary\n\nFCD dataset### Supported Tasks and Leaderboards\n\nNLP### Languages\n\nen## Dataset Structure### Data Instances### Data Fields### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data### Social Impact of Dataset### Discussion of Biases### Other Known Limitations## Additional Information### Dataset Curators### Licensing Information" ]

[ "license:other", "arxiv:2108.05289", "region:us" ]

[ "2108.05289" ]

[]

[]

[ "TAGS\n#license-other #arxiv-2108.05289 #region-us \n" ]

[ 19 ]

[ "passage: TAGS\n#license-other #arxiv-2108.05289 #region-us \n" ]

[ "task_categories:text-classification", "task_categories:text2text-generation", "task_ids:natural-language-inference", "task_ids:explanation-generation", "annotations_creators:expert-generated", "language_creators:expert-generated", "language_creators:machine-generated", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:1K<n<10K", "source_datasets:original", "language:en", "license:afl-3.0", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for FigLang2022SharedTask", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: TBA\n- Point of Contact: URL@URL", "### Dataset Summary\n\nModel in the loop approach for fig lang generation and explainability", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\n\n\n\n\nTBA", "### Contributions\n\nThanks to @github-username for adding this dataset." ]

[ "TAGS\n#task_categories-text-classification #task_categories-text2text-generation #task_ids-natural-language-inference #task_ids-explanation-generation #annotations_creators-expert-generated #language_creators-expert-generated #language_creators-machine-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-English #license-afl-3.0 #region-us \n", "# Dataset Card for FigLang2022SharedTask", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: TBA\n- Point of Contact: URL@URL", "### Dataset Summary\n\nModel in the loop approach for fig lang generation and explainability", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\n\n\n\n\nTBA", "### Contributions\n\nThanks to @github-username for adding this dataset." ]

[ 141, 13, 113, 26, 19, 6, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 8, 19 ]

[ "passage: TAGS\n#task_categories-text-classification #task_categories-text2text-generation #task_ids-natural-language-inference #task_ids-explanation-generation #annotations_creators-expert-generated #language_creators-expert-generated #language_creators-machine-generated #language_creators-crowdsourced #multilinguality-monolingual #size_categories-1K<n<10K #source_datasets-original #language-English #license-afl-3.0 #region-us \n# Dataset Card for FigLang2022SharedTask## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: TBA\n- Point of Contact: URL@URL### Dataset Summary\n\nModel in the loop approach for fig lang generation and explainability## Dataset Structure### Data Instances### Data Fields### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data### Social Impact of Dataset### Discussion of Biases### Other Known Limitations## Additional Information### Dataset Curators### Licensing Information\n\n\n\n\n\nTBA### Contributions\n\nThanks to @github-username for adding this dataset." ]

[ "task_categories:translation", "annotations_creators:found", "language_creators:found", "multilinguality:translation", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "language:pt", "license:afl-3.0", "region:us" ]

[]

[ "en", "pt" ]

[]

[ "TAGS\n#task_categories-translation #annotations_creators-found #language_creators-found #multilinguality-translation #size_categories-100K<n<1M #source_datasets-original #language-English #language-Portuguese #license-afl-3.0 #region-us \n" ]

[ 78 ]

[ "passage: TAGS\n#task_categories-translation #annotations_creators-found #language_creators-found #multilinguality-translation #size_categories-100K<n<1M #source_datasets-original #language-English #language-Portuguese #license-afl-3.0 #region-us \n" ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Question Answering\n* Model: autoevaluate/extractive-question-answering\n* Dataset: squad\n* Config: plain_text\n* Split: validation\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @None for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Question Answering\n* Model: autoevaluate/extractive-question-answering\n* Dataset: squad\n* Config: plain_text\n* Split: validation\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @None for evaluating this model." ]

[ 13, 91, 15 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Question Answering\n* Model: autoevaluate/extractive-question-answering\n* Dataset: squad\n* Config: plain_text\n* Split: validation\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @None for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: lewtun/autotrain-acronym-identification-7324788\n* Dataset: acronym_identification\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @Felixver for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: lewtun/autotrain-acronym-identification-7324788\n* Dataset: acronym_identification\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @Felixver for evaluating this model." ]

[ 13, 87, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: lewtun/autotrain-acronym-identification-7324788\n* Dataset: acronym_identification\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @Felixver for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: AlexanderPeter/bert-finetuned-ner\n* Dataset: conll2003\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @test for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: AlexanderPeter/bert-finetuned-ner\n* Dataset: conll2003\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @test for evaluating this model." ]

[ 13, 78, 14 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: AlexanderPeter/bert-finetuned-ner\n* Dataset: conll2003\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @test for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: ArBert/roberta-base-finetuned-ner-kmeans\n* Dataset: conll2003\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @test for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: ArBert/roberta-base-finetuned-ner-kmeans\n* Dataset: conll2003\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @test for evaluating this model." ]

[ 13, 86, 14 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Token Classification\n* Model: ArBert/roberta-base-finetuned-ner-kmeans\n* Dataset: conll2003\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @test for evaluating this model." ]

[ "region:us" ]

[]

[]

[ "## Dataset Description\n\n- Homepage:\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nSample dataset for the Library of Congress Historical American Buildings collection.", "### Supported Tasks and Leaderboards", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances", "### Data Fields\n\n'call_number'\n'control_number'\n'created'\n'created_published'\n'created_published_date'\n'creators'\n'date'\n'display_offsite'\n'id'\n'link'\n'medium_brief'\n'mediums'\n'modified'\n'notes'\n'part_of'\n'part_of_group'\n'place'\n'latitude'\n'link'\n'longitude'\n'title'\n'repository'\n'resource_links'\n'rights_advisory'\n'rights_information'\n'service_low'\n'service_medium'\n'source_created'\n'source_modified'\n'subject_headings'\n'thumb_gallery'\n'title'", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\ncc0-1.0" ]

[ "TAGS\n#region-us \n", "## Dataset Description\n\n- Homepage:\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nSample dataset for the Library of Congress Historical American Buildings collection.", "### Supported Tasks and Leaderboards", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances", "### Data Fields\n\n'call_number'\n'control_number'\n'created'\n'created_published'\n'created_published_date'\n'creators'\n'date'\n'display_offsite'\n'id'\n'link'\n'medium_brief'\n'mediums'\n'modified'\n'notes'\n'part_of'\n'part_of_group'\n'place'\n'latitude'\n'link'\n'longitude'\n'title'\n'repository'\n'resource_links'\n'rights_advisory'\n'rights_information'\n'service_low'\n'service_medium'\n'source_created'\n'source_modified'\n'subject_headings'\n'thumb_gallery'\n'title'", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\ncc0-1.0" ]

[ 6, 24, 22, 10, 5, 6, 6, 171, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 5, 6, 11 ]

[ "passage: TAGS\n#region-us \n## Dataset Description\n\n- Homepage:\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:### Dataset Summary\n\nSample dataset for the Library of Congress Historical American Buildings collection.### Supported Tasks and Leaderboards### Languages\n\nEnglish## Dataset Structure### Data Instances### Data Fields\n\n'call_number'\n'control_number'\n'created'\n'created_published'\n'created_published_date'\n'creators'\n'date'\n'display_offsite'\n'id'\n'link'\n'medium_brief'\n'mediums'\n'modified'\n'notes'\n'part_of'\n'part_of_group'\n'place'\n'latitude'\n'link'\n'longitude'\n'title'\n'repository'\n'resource_links'\n'rights_advisory'\n'rights_information'\n'service_low'\n'service_medium'\n'source_created'\n'source_modified'\n'subject_headings'\n'thumb_gallery'\n'title'### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data## Additional Information### Dataset Curators### Licensing Information\n\ncc0-1.0" ]

[ "generated_from_keras_callback", "region:us" ]

[]

[]

[ "# ms-en\n\nNotebooks to gather the dataset at URL\n\nPreparation notebook at URL" ]

[ "TAGS\n#generated_from_keras_callback #region-us \n", "# ms-en\n\nNotebooks to gather the dataset at URL\n\nPreparation notebook at URL" ]

[ 17, 19 ]

[ "passage: TAGS\n#generated_from_keras_callback #region-us \n# ms-en\n\nNotebooks to gather the dataset at URL\n\nPreparation notebook at URL" ]

[ "generated_from_keras_callback", "region:us" ]

[]

[]

[ "# en-ms\n\nNotebooks to gather the dataset at URL\n\nPreparation notebook at URL" ]

[ "TAGS\n#generated_from_keras_callback #region-us \n", "# en-ms\n\nNotebooks to gather the dataset at URL\n\nPreparation notebook at URL" ]

[ 17, 19 ]

[ "passage: TAGS\n#generated_from_keras_callback #region-us \n# en-ms\n\nNotebooks to gather the dataset at URL\n\nPreparation notebook at URL" ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-cnn_dailymail\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @lewtun for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-cnn_dailymail\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @lewtun for evaluating this model." ]

[ 13, 77, 15 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-cnn_dailymail\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @lewtun for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/distilbert-base-uncased-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/distilbert-base-uncased-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ 13, 87, 19 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/distilbert-base-uncased-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/roberta-base-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/roberta-base-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ 13, 82, 19 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/roberta-base-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/bert-base-uncased-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/bert-base-uncased-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ 13, 85, 19 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Multi-class Text Classification\n* Model: bhadresh-savani/bert-base-uncased-emotion\n* Dataset: emotion\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @bhadresh-savani for evaluating this model." ]

[ "task_categories:text-generation", "task_ids:language-modeling", "language_creators:crowdsourced", "language_creators:expert-generated", "multilinguality:muonolingual", "size_categories:unknown", "language:code", "license:other", "region:us" ]

[]

[ "code" ]

[ "# GitHub Jupyter Dataset", "## Dataset Description\nThis is a parsed and preprocessed version of GitHub-Jupyter Dataset, a dataset extracted from Jupyter Notebooks on BigQuery. We only keep markdown and python cells and convert the markdown to text. Some heuristics are also applied to filter notebooks with little data and very long or very short cells.", "## Licenses\nEach example has the license of its associated repository. There are in total 15 licenses:" ]

[ "TAGS\n#task_categories-text-generation #task_ids-language-modeling #language_creators-crowdsourced #language_creators-expert-generated #multilinguality-muonolingual #size_categories-unknown #language-code #license-other #region-us \n", "# GitHub Jupyter Dataset", "## Dataset Description\nThis is a parsed and preprocessed version of GitHub-Jupyter Dataset, a dataset extracted from Jupyter Notebooks on BigQuery. We only keep markdown and python cells and convert the markdown to text. Some heuristics are also applied to filter notebooks with little data and very long or very short cells.", "## Licenses\nEach example has the license of its associated repository. There are in total 15 licenses:" ]

[ 76, 9, 82, 23 ]

[ "passage: TAGS\n#task_categories-text-generation #task_ids-language-modeling #language_creators-crowdsourced #language_creators-expert-generated #multilinguality-muonolingual #size_categories-unknown #language-code #license-other #region-us \n# GitHub Jupyter Dataset## Dataset Description\nThis is a parsed and preprocessed version of GitHub-Jupyter Dataset, a dataset extracted from Jupyter Notebooks on BigQuery. We only keep markdown and python cells and convert the markdown to text. Some heuristics are also applied to filter notebooks with little data and very long or very short cells.## Licenses\nEach example has the license of its associated repository. There are in total 15 licenses:" ]

[ "task_categories:question-answering", "task_ids:open-domain-qa", "annotations_creators:no-annotation", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:cc-by-sa-3.0", "region:us" ]

[]

[ "en" ]

[ "### Dataset Summary\n\n\nThe NQ corpus contains questions from real users, and it requires QA systems to\nread and comprehend an entire Wikipedia article that may or may not contain the\nanswer to the question. The inclusion of real user questions, and the\nrequirement that solutions should read an entire page to find the answer, cause\nNQ to be a more realistic and challenging task than prior QA datasets.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### default\n\n\n* Size of downloaded dataset files: 42981 MB\n* Size of the generated dataset: 139706 MB\n* Total amount of disk used: 182687 MB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.", "#### default", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nCreative Commons Attribution-ShareAlike 3.0 Unported.", "### Contributions" ]

[ "TAGS\n#task_categories-question-answering #task_ids-open-domain-qa #annotations_creators-no-annotation #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-cc-by-sa-3.0 #region-us \n", "### Dataset Summary\n\n\nThe NQ corpus contains questions from real users, and it requires QA systems to\nread and comprehend an entire Wikipedia article that may or may not contain the\nanswer to the question. The inclusion of real user questions, and the\nrequirement that solutions should read an entire page to find the answer, cause\nNQ to be a more realistic and challenging task than prior QA datasets.", "### Supported Tasks and Leaderboards", "### Languages\n\n\nDataset Structure\n-----------------", "### Data Instances", "#### default\n\n\n* Size of downloaded dataset files: 42981 MB\n* Size of the generated dataset: 139706 MB\n* Total amount of disk used: 182687 MB\n\n\nAn example of 'train' looks as follows.", "### Data Fields\n\n\nThe data fields are the same among all splits.", "#### default", "### Data Splits\n\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators", "### Licensing Information\n\n\nCreative Commons Attribution-ShareAlike 3.0 Unported.", "### Contributions" ]

[ 97, 87, 10, 11, 6, 49, 17, 3, 11, 7, 4, 10, 10, 5, 5, 9, 18, 7, 8, 14, 6, 14, 5 ]

[ "passage: TAGS\n#task_categories-question-answering #task_ids-open-domain-qa #annotations_creators-no-annotation #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-cc-by-sa-3.0 #region-us \n### Dataset Summary\n\n\nThe NQ corpus contains questions from real users, and it requires QA systems to\nread and comprehend an entire Wikipedia article that may or may not contain the\nanswer to the question. The inclusion of real user questions, and the\nrequirement that solutions should read an entire page to find the answer, cause\nNQ to be a more realistic and challenging task than prior QA datasets.### Supported Tasks and Leaderboards### Languages\n\n\nDataset Structure\n-----------------### Data Instances#### default\n\n\n* Size of downloaded dataset files: 42981 MB\n* Size of the generated dataset: 139706 MB\n* Total amount of disk used: 182687 MB\n\n\nAn example of 'train' looks as follows.### Data Fields\n\n\nThe data fields are the same among all splits.#### default### Data Splits\n\n\n\nDataset Creation\n----------------### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------### Social Impact of Dataset### Discussion of Biases### Other Known Limitations\n\n\nAdditional Information\n----------------------### Dataset Curators### Licensing Information\n\n\nCreative Commons Attribution-ShareAlike 3.0 Unported.### Contributions" ]

[ "task_categories:graph-ml", "license:mit", "region:us" ]

[]

[]

[ "### Dataset Summary\n\n\nThe 'ogbg-molhiv' dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.", "### Supported Tasks and Leaderboards\n\n\n'ogbg-molhiv' should be used for molecular property prediction (aiming to predict whether molecules inhibit HIV or not), a binary classification task. The score used is ROC-AUC.\n\n\nThe associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'node\\_feat' (list: #nodes x #node-features): nodes\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)\n* 'num\\_nodes' (int): number of nodes of the graph", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under MIT license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "TAGS\n#task_categories-graph-ml #license-mit #region-us \n", "### Dataset Summary\n\n\nThe 'ogbg-molhiv' dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.", "### Supported Tasks and Leaderboards\n\n\n'ogbg-molhiv' should be used for molecular property prediction (aiming to predict whether molecules inhibit HIV or not), a binary classification task. The score used is ROC-AUC.\n\n\nThe associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'node\\_feat' (list: #nodes x #node-features): nodes\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)\n* 'num\\_nodes' (int): number of nodes of the graph", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under MIT license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ 21, 51, 82, 25, 4, 158, 47, 16, 17 ]

[ "passage: TAGS\n#task_categories-graph-ml #license-mit #region-us \n### Dataset Summary\n\n\nThe 'ogbg-molhiv' dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.### Supported Tasks and Leaderboards\n\n\n'ogbg-molhiv' should be used for molecular property prediction (aiming to predict whether molecules inhibit HIV or not), a binary classification task. The score used is ROC-AUC.\n\n\nThe associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.\n\n\nExternal Use\n------------### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------### Data Properties### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'node\\_feat' (list: #nodes x #node-features): nodes\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)\n* 'num\\_nodes' (int): number of nodes of the graph### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------### Licensing Information\n\n\nThe dataset has been released under MIT license.### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "license:cc-by-sa-4.0", "arxiv:2206.08219", "region:us" ]

[ "2206.08219" ]

[]

[ "### Dataset Summary\n\n\n for hand gesture recognition (HGR) systems. You can use it for image classification or image detection tasks. Proposed dataset allows to build HGR systems, which can be used in video conferencing services (Zoom, Skype, Discord, Jazz etc.), home automation systems, the automotive sector, etc.\n\n\nHaGRID size is 716GB and dataset contains 552,992 FullHD (1920 × 1080) RGB images divided into 18 classes of gestures. Also, some images have 'no\\_gesture' class if there is a second free hand in the frame. This extra class contains 123,589 samples. The data were split into training 92%, and testing 8% sets by subject user-id, with 509,323 images for train and 43,669 images for test.\n\n\n and 'leading\\_conf' as confidence for 'leading\\_hand' annotation. We provide 'user\\_id' field that will allow you to split the train / val dataset yourself.\n\n\nDownloads\n---------\n\n\nWe split the train dataset into 18 archives by gestures because of the large size of data. Download and unzip them from the following links:", "### Trainval\n\n\n\n'train\\_val' annotations: 'ann\\_train\\_val'", "### Test\n\n\nTest: images, Archives: 'test', Size: 60.4 GB\nTest: annotations, Archives: 'ann\\_test', Size: 3.4 MB", "### Subsample\n\n\nSubsample has 100 items per gesture.\n\n\nSubsample: images, Archives: 'subsample', Size: 2.5 GB\nSubsample: annotations, Archives: 'ann\\_subsample', Size: 153.8 KB\n\n\nModels\n------\n\n\nWe provide some pre-trained classifiers and one detector as baselines.\n\n\nClassifiers: ResNet18, F1 Gesture: 98.72, F1 Leading hand: 99.27\nClassifiers: ResNet152, F1 Gesture: 99.11, F1 Leading hand: 99.45\nClassifiers: ResNeXt50, F1 Gesture: 98.99, F1 Leading hand: 99.39\nClassifiers: ResNeXt101, F1 Gesture: 99.28, F1 Leading hand: 99.28\nClassifiers: MobileNetV3-small, F1 Gesture: 96.78, F1 Leading hand: 98.28\nClassifiers: MobileNetV3-large, F1 Gesture: 97.88, F1 Leading hand: 98.58\nClassifiers: VitB-32, F1 Gesture: 98.49, F1 Leading hand: 99.13\n\n\n\nLinks\n-----\n\n\n* Github, Mirror\n* arXiv\n* Paperswithcode", "### Supported Tasks and Leaderboards", "### Languages\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields", "### Data Splits\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThis dataset was shared by @kapitanov", "### Licensing Information\n\n\nThe license for this dataset is cc-by-sa-4.0", "### Contributions" ]

[ "TAGS\n#license-cc-by-sa-4.0 #arxiv-2206.08219 #region-us \n", "### Dataset Summary\n\n\n for hand gesture recognition (HGR) systems. You can use it for image classification or image detection tasks. Proposed dataset allows to build HGR systems, which can be used in video conferencing services (Zoom, Skype, Discord, Jazz etc.), home automation systems, the automotive sector, etc.\n\n\nHaGRID size is 716GB and dataset contains 552,992 FullHD (1920 × 1080) RGB images divided into 18 classes of gestures. Also, some images have 'no\\_gesture' class if there is a second free hand in the frame. This extra class contains 123,589 samples. The data were split into training 92%, and testing 8% sets by subject user-id, with 509,323 images for train and 43,669 images for test.\n\n\n and 'leading\\_conf' as confidence for 'leading\\_hand' annotation. We provide 'user\\_id' field that will allow you to split the train / val dataset yourself.\n\n\nDownloads\n---------\n\n\nWe split the train dataset into 18 archives by gestures because of the large size of data. Download and unzip them from the following links:", "### Trainval\n\n\n\n'train\\_val' annotations: 'ann\\_train\\_val'", "### Test\n\n\nTest: images, Archives: 'test', Size: 60.4 GB\nTest: annotations, Archives: 'ann\\_test', Size: 3.4 MB", "### Subsample\n\n\nSubsample has 100 items per gesture.\n\n\nSubsample: images, Archives: 'subsample', Size: 2.5 GB\nSubsample: annotations, Archives: 'ann\\_subsample', Size: 153.8 KB\n\n\nModels\n------\n\n\nWe provide some pre-trained classifiers and one detector as baselines.\n\n\nClassifiers: ResNet18, F1 Gesture: 98.72, F1 Leading hand: 99.27\nClassifiers: ResNet152, F1 Gesture: 99.11, F1 Leading hand: 99.45\nClassifiers: ResNeXt50, F1 Gesture: 98.99, F1 Leading hand: 99.39\nClassifiers: ResNeXt101, F1 Gesture: 99.28, F1 Leading hand: 99.28\nClassifiers: MobileNetV3-small, F1 Gesture: 96.78, F1 Leading hand: 98.28\nClassifiers: MobileNetV3-large, F1 Gesture: 97.88, F1 Leading hand: 98.58\nClassifiers: VitB-32, F1 Gesture: 98.49, F1 Leading hand: 99.13\n\n\n\nLinks\n-----\n\n\n* Github, Mirror\n* arXiv\n* Paperswithcode", "### Supported Tasks and Leaderboards", "### Languages\n\n\nDataset Structure\n-----------------", "### Data Instances", "### Data Fields", "### Data Splits\n\n\nDataset Creation\n----------------", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\n\nAdditional Information\n----------------------", "### Dataset Curators\n\n\nThis dataset was shared by @kapitanov", "### Licensing Information\n\n\nThe license for this dataset is cc-by-sa-4.0", "### Contributions" ]

[ 26, 472, 25, 37, 289, 10, 11, 6, 5, 11, 7, 4, 10, 10, 5, 5, 9, 18, 7, 8, 14, 16, 21, 5 ]

[ "passage: TAGS\n#license-cc-by-sa-4.0 #arxiv-2206.08219 #region-us \n### Dataset Summary\n\n\n for hand gesture recognition (HGR) systems. You can use it for image classification or image detection tasks. Proposed dataset allows to build HGR systems, which can be used in video conferencing services (Zoom, Skype, Discord, Jazz etc.), home automation systems, the automotive sector, etc.\n\n\nHaGRID size is 716GB and dataset contains 552,992 FullHD (1920 × 1080) RGB images divided into 18 classes of gestures. Also, some images have 'no\\_gesture' class if there is a second free hand in the frame. This extra class contains 123,589 samples. The data were split into training 92%, and testing 8% sets by subject user-id, with 509,323 images for train and 43,669 images for test.\n\n\n and 'leading\\_conf' as confidence for 'leading\\_hand' annotation. We provide 'user\\_id' field that will allow you to split the train / val dataset yourself.\n\n\nDownloads\n---------\n\n\nWe split the train dataset into 18 archives by gestures because of the large size of data. Download and unzip them from the following links:", "passage: ### Trainval\n\n\n\n'train\\_val' annotations: 'ann\\_train\\_val'### Test\n\n\nTest: images, Archives: 'test', Size: 60.4 GB\nTest: annotations, Archives: 'ann\\_test', Size: 3.4 MB### Subsample\n\n\nSubsample has 100 items per gesture.\n\n\nSubsample: images, Archives: 'subsample', Size: 2.5 GB\nSubsample: annotations, Archives: 'ann\\_subsample', Size: 153.8 KB\n\n\nModels\n------\n\n\nWe provide some pre-trained classifiers and one detector as baselines.\n\n\nClassifiers: ResNet18, F1 Gesture: 98.72, F1 Leading hand: 99.27\nClassifiers: ResNet152, F1 Gesture: 99.11, F1 Leading hand: 99.45\nClassifiers: ResNeXt50, F1 Gesture: 98.99, F1 Leading hand: 99.39\nClassifiers: ResNeXt101, F1 Gesture: 99.28, F1 Leading hand: 99.28\nClassifiers: MobileNetV3-small, F1 Gesture: 96.78, F1 Leading hand: 98.28\nClassifiers: MobileNetV3-large, F1 Gesture: 97.88, F1 Leading hand: 98.58\nClassifiers: VitB-32, F1 Gesture: 98.49, F1 Leading hand: 99.13\n\n\n\nLinks\n-----\n\n\n* Github, Mirror\n* arXiv\n* Paperswithcode### Supported Tasks and Leaderboards### Languages\n\n\nDataset Structure\n-----------------### Data Instances### Data Fields### Data Splits\n\n\nDataset Creation\n----------------### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information\n\n\nConsiderations for Using the Data\n---------------------------------### Social Impact of Dataset### Discussion of Biases### Other Known Limitations\n\n\nAdditional Information\n----------------------### Dataset Curators\n\n\nThis dataset was shared by @kapitanov### Licensing Information\n\n\nThe license for this dataset is cc-by-sa-4.0" ]

[ "task_categories:graph-ml", "license:mit", "region:us" ]

[]

[]

[ "### Dataset Summary\n\n\nThe 'ogbg-molpcba' dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.", "### Supported Tasks and Leaderboards\n\n\n'ogbg-molpcba' should be used for molecular property prediction (with 128 properties to predict, not all present for all graphs), a binary classification task.\nThe score used is Average Precision (AP) averaged over the tasks.\n\n\nThe associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'node\\_feat' (list: #nodes x #node-features): nodes\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 128 labels, equal to zero, one, or Nan if the property is not relevant for the graph)\n* 'num\\_nodes' (int): number of nodes of the graph", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under MIT license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "TAGS\n#task_categories-graph-ml #license-mit #region-us \n", "### Dataset Summary\n\n\nThe 'ogbg-molpcba' dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.", "### Supported Tasks and Leaderboards\n\n\n'ogbg-molpcba' should be used for molecular property prediction (with 128 properties to predict, not all present for all graphs), a binary classification task.\nThe score used is Average Precision (AP) averaged over the tasks.\n\n\nThe associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'node\\_feat' (list: #nodes x #node-features): nodes\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 128 labels, equal to zero, one, or Nan if the property is not relevant for the graph)\n* 'num\\_nodes' (int): number of nodes of the graph", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under MIT license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ 21, 51, 92, 25, 4, 173, 47, 16, 17 ]

[ "passage: TAGS\n#task_categories-graph-ml #license-mit #region-us \n### Dataset Summary\n\n\nThe 'ogbg-molpcba' dataset is a small molecular property prediction dataset, adapted from MoleculeNet by teams at Stanford, to be a part of the Open Graph Benchmark.### Supported Tasks and Leaderboards\n\n\n'ogbg-molpcba' should be used for molecular property prediction (with 128 properties to predict, not all present for all graphs), a binary classification task.\nThe score used is Average Precision (AP) averaged over the tasks.\n\n\nThe associated leaderboards are here: OGB leaderboard and Papers with code leaderboard.\n\n\nExternal Use\n------------### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------### Data Properties### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'node\\_feat' (list: #nodes x #node-features): nodes\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 128 labels, equal to zero, one, or Nan if the property is not relevant for the graph)\n* 'num\\_nodes' (int): number of nodes of the graph### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------### Licensing Information\n\n\nThe dataset has been released under MIT license.### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-5k\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-5k\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 27, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-5k\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ 6 ]

[ "passage: TAGS\n#region-us \n" ]

[ "region:us" ]

[]

[]

[ "### Languages\n\n\nThe BCP-47 code for the dataset's language is unk.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nA sample from this dataset looks as follows:", "### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):", "### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ "TAGS\n#region-us \n", "### Languages\n\n\nThe BCP-47 code for the dataset's language is unk.\n\n\nDataset Structure\n-----------------", "### Data Instances\n\n\nA sample from this dataset looks as follows:", "### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):", "### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ 6, 27, 17, 23, 27 ]

[ "passage: TAGS\n#region-us \n### Languages\n\n\nThe BCP-47 code for the dataset's language is unk.\n\n\nDataset Structure\n-----------------### Data Instances\n\n\nA sample from this dataset looks as follows:### Dataset Fields\n\n\nThe dataset has the following fields (also called \"features\"):### Dataset Splits\n\n\nThis dataset is split into a train and validation split. The split sizes are as follow:" ]

[ "region:us" ]

[]

[]

[ "# K-12Corpus" ]

[ "TAGS\n#region-us \n", "# K-12Corpus" ]

[ 6, 5 ]

[ "passage: TAGS\n#region-us \n# K-12Corpus" ]

[ "region:us" ]

[]

[]

[]

[ "TAGS\n#region-us \n" ]

[ 6 ]

[ "passage: TAGS\n#region-us \n" ]

[ "task_categories:graph-ml", "license:cc0-1.0", "region:us" ]

[]

[]

[ "### Dataset Summary\n\n\nThe 'ogbg-ppa' dataset is \"a set of undirected protein association neighborhoods extracted from the protein-protein association networks of 1,581 species\", over 37 taxonomic groups, by teams at Stanford, to be a part of the Open Graph Benchmark. See their website for dataset postprocessing.", "### Supported Tasks and Leaderboards\n\n\n'ogbg-ppa' should be used for taxonomic group prediction, a 37-way multi-class classification task. The score used is Average Precision on the test set.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)\n* 'num\\_nodes' (int): number of nodes of the graph\n\n\nThe nodes don't have specific features and are implicit from the lists of edges", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under CC-0 license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "TAGS\n#task_categories-graph-ml #license-cc0-1.0 #region-us \n", "### Dataset Summary\n\n\nThe 'ogbg-ppa' dataset is \"a set of undirected protein association neighborhoods extracted from the protein-protein association networks of 1,581 species\", over 37 taxonomic groups, by teams at Stanford, to be a part of the Open Graph Benchmark. See their website for dataset postprocessing.", "### Supported Tasks and Leaderboards\n\n\n'ogbg-ppa' should be used for taxonomic group prediction, a 37-way multi-class classification task. The score used is Average Precision on the test set.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)\n* 'num\\_nodes' (int): number of nodes of the graph\n\n\nThe nodes don't have specific features and are implicit from the lists of edges", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under CC-0 license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ 24, 78, 56, 25, 4, 153, 47, 17, 17 ]

[ "passage: TAGS\n#task_categories-graph-ml #license-cc0-1.0 #region-us \n### Dataset Summary\n\n\nThe 'ogbg-ppa' dataset is \"a set of undirected protein association neighborhoods extracted from the protein-protein association networks of 1,581 species\", over 37 taxonomic groups, by teams at Stanford, to be a part of the Open Graph Benchmark. See their website for dataset postprocessing.### Supported Tasks and Leaderboards\n\n\n'ogbg-ppa' should be used for taxonomic group prediction, a 37-way multi-class classification task. The score used is Average Precision on the test set.\n\n\nExternal Use\n------------### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------### Data Properties### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_attr' (list: #edges x #edge-features): for the aforementioned edges, contains their features\n* 'y' (list: 1 x #labels): contains the number of labels available to predict (here 1, equal to zero or one)\n* 'num\\_nodes' (int): number of nodes of the graph\n\n\nThe nodes don't have specific features and are implicit from the lists of edges### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nAdditional Information\n----------------------### Licensing Information\n\n\nThe dataset has been released under CC-0 license.### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "task_categories:graph-ml", "license:mit", "region:us" ]

[]

[]

[ "### Dataset Summary\n\n\nThe 'ogbg-code2' dataset contains Abstract Syntax Trees (ASTs) obtained from 450 thousands Python method definitions, from GitHub CodeSearchNet. \"Methods are extracted from a total of 13,587 different repositories across the most popular projects on GitHub.\", by teams at Stanford, to be a part of the Open Graph Benchmark. See their website or paper for dataset postprocessing.", "### Supported Tasks and Leaderboards\n\n\n\"The task is to predict the sub-tokens forming the method name, given the Python method body represented by AST and its node features. This task is often referred to as “code summarization”, because the model is trained to find succinct and precise description for a complete logical unit.\"\n\n\nThe score is the F1 score of sub-token prediction.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_feat' (list: #edges x #edge-features): features of edges\n* 'node\\_feat' (list: #nodes x #node-features): the nodes features, embedded\n* 'node\\_feat\\_expanded' (list: #nodes x #node-features): the nodes features, as code\n* 'node\\_is\\_attributed' (list: 1 x #nodes): ?\n* 'node\\_dfs\\_order' (list: #nodes x #1): the nodes order in the abstract tree, if parsed using a depth first search\n* 'node\\_depth' (list: #nodes x #1): the nodes depth in the abstract tree\n* 'y' (list: 1 x #tokens): contains the tokens to predict as method name\n* 'num\\_nodes' (int): number of nodes of the graph\n* 'ptr' (list: 2): index of first and last node of the graph\n* 'batch' (list: 1 x #nodes): ?", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nMore information ('node\\_feat\\_expanded') has been added through the typeidx2type and attridx2attr csv files of the repo.\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under MIT license license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ "TAGS\n#task_categories-graph-ml #license-mit #region-us \n", "### Dataset Summary\n\n\nThe 'ogbg-code2' dataset contains Abstract Syntax Trees (ASTs) obtained from 450 thousands Python method definitions, from GitHub CodeSearchNet. \"Methods are extracted from a total of 13,587 different repositories across the most popular projects on GitHub.\", by teams at Stanford, to be a part of the Open Graph Benchmark. See their website or paper for dataset postprocessing.", "### Supported Tasks and Leaderboards\n\n\n\"The task is to predict the sub-tokens forming the method name, given the Python method body represented by AST and its node features. This task is often referred to as “code summarization”, because the model is trained to find succinct and precise description for a complete logical unit.\"\n\n\nThe score is the F1 score of sub-token prediction.\n\n\nExternal Use\n------------", "### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------", "### Data Properties", "### Data Fields\n\n\nEach row of a given file is a graph, with:\n\n\n* 'edge\\_index' (list: 2 x #edges): pairs of nodes constituting edges\n* 'edge\\_feat' (list: #edges x #edge-features): features of edges\n* 'node\\_feat' (list: #nodes x #node-features): the nodes features, embedded\n* 'node\\_feat\\_expanded' (list: #nodes x #node-features): the nodes features, as code\n* 'node\\_is\\_attributed' (list: 1 x #nodes): ?\n* 'node\\_dfs\\_order' (list: #nodes x #1): the nodes order in the abstract tree, if parsed using a depth first search\n* 'node\\_depth' (list: #nodes x #1): the nodes depth in the abstract tree\n* 'y' (list: 1 x #tokens): contains the tokens to predict as method name\n* 'num\\_nodes' (int): number of nodes of the graph\n* 'ptr' (list: 2): index of first and last node of the graph\n* 'batch' (list: 1 x #nodes): ?", "### Data Splits\n\n\nThis data comes from the PyGeometric version of the dataset provided by OGB, and follows the provided data splits.\nThis information can be found back using\n\n\nMore information ('node\\_feat\\_expanded') has been added through the typeidx2type and attridx2attr csv files of the repo.\n\n\nAdditional Information\n----------------------", "### Licensing Information\n\n\nThe dataset has been released under MIT license license.", "### Contributions\n\n\nThanks to @clefourrier for adding this dataset." ]

[ 21, 103, 98, 25, 4, 301, 87, 17, 17 ]

[ "passage: TAGS\n#task_categories-graph-ml #license-mit #region-us \n### Dataset Summary\n\n\nThe 'ogbg-code2' dataset contains Abstract Syntax Trees (ASTs) obtained from 450 thousands Python method definitions, from GitHub CodeSearchNet. \"Methods are extracted from a total of 13,587 different repositories across the most popular projects on GitHub.\", by teams at Stanford, to be a part of the Open Graph Benchmark. See their website or paper for dataset postprocessing.### Supported Tasks and Leaderboards\n\n\n\"The task is to predict the sub-tokens forming the method name, given the Python method body represented by AST and its node features. This task is often referred to as “code summarization”, because the model is trained to find succinct and precise description for a complete logical unit.\"\n\n\nThe score is the F1 score of sub-token prediction.\n\n\nExternal Use\n------------### PyGeometric\n\n\nTo load in PyGeometric, do the following:\n\n\nDataset Structure\n-----------------### Data Properties" ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-base\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @sunilmallya for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-base\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @sunilmallya for evaluating this model." ]

[ 13, 69, 17 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-base\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @sunilmallya for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: philschmid/distilbart-cnn-12-6-samsum\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ArjunPrSarkhel for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: philschmid/distilbart-cnn-12-6-samsum\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ArjunPrSarkhel for evaluating this model." ]

[ 13, 82, 20 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: philschmid/distilbart-cnn-12-6-samsum\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ArjunPrSarkhel for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: eslamxm/mbert2mbert-finetune-fa\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @iserralv for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: eslamxm/mbert2mbert-finetune-fa\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @iserralv for evaluating this model." ]

[ 13, 81, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: eslamxm/mbert2mbert-finetune-fa\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @iserralv for evaluating this model." ]

[ "task_categories:text-classification", "task_ids:acceptability-classification", "annotations_creators:machine-generated", "annotations_creators:expert-generated", "language_creators:expert-generated", "multilinguality:monolingual", "size_categories:10K<n<100K", "source_datasets:original", "language:it", "license:cc-by-4.0", "region:us" ]

[]

[ "it" ]

[ "# Dataset Card for Annotated dataset to assess the accuracy of the textual description of cultural heritage records", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage:URL\n- Repository:URL\n- Paper:URL\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nThe dataset contains more than 100K textual descriptions of cultural items from Cultura Italia, the Italian National Cultural aggregator. Each of the description is labeled either HIGH or LOW quality, according its adherence to the standard cataloguing guidelines provided by Istituto Centrale per il Catalogo e la Documentazione (ICCD). More precisely, each description is labeled as HIGH quality if the object and subject of the item (for which the description is provided) are both described according to the ICCD guidelines, and as LOW quality in all other cases. Most of the dataset was manually annotated, with ~30K descriptions automatically labeled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections. The dataset was developed to support the training and testing of ML text classification approaches for automatically assessing the quality of textual descriptions in digital Cultural Heritage repositories.", "### Supported Tasks and Leaderboards\n\nThis dataset can be used for text classification tasks. The paper introducing the dataset achieved an f1 score of '.783' for the task of classifying if a metadata record was low or high quality. Please see the results table for a full overview of the results reported in the paper.", "### Languages\n\nThe dataset consists of Italian metadata records. The labels are in English.", "## Dataset Structure\n\nThe dataset has only one configuration.", "### Data Instances\n\nAn example instance from the dataset:", "### Data Fields\n\nThe datafields are:\n\n- 'metadata_text': this contains the metadata text which was sourced from Cultura Italia\n- 'label': this is the label indicating if the record is 'High_Quality', or 'Low_Quality'. Most of the dataset was manually annotated, with ~30K descriptions automatically labelled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections.\n- 'source': the source of the metadata record", "### Data Splits\n\nThe dataset used 'ten-fold cross-validation' and doesn't report specific splits for train, validation and test data.", "## Dataset Creation\n\nThe dataset was generated using records from Cultura Italia. From the paper introducing the dataset:\n\n> By using the textual description encoded by the dc:description element from the Dublin Core metadata schema, we collect a dataset of 100,821 descriptions, after duplicate removal. These records include mainly data from “Musei d’Italia” and “Regione Marche” datasets, which have been chosen because they contain a high number of non-empty dc:description elements. p.221", "### Curation Rationale\n\nFrom the paper:\n\n> Duplicates were removed for two reasons: this reduced annotation effort in the subsequent manual annotation, and avoided that the same example appear both in the training and in the test set, a situation that could make classification biased and lead to inaccurate evaluation in supervised settings.Footnote 10 Duplicated descriptions were mainly short and of low-quality, reporting few generic words to describe an item (e.g. “Mensola.”, “Dipinto.”). p.221", "### Source Data", "#### Initial Data Collection and Normalization\n\nThe dataset was generated using records from Cultura Italia. This repository is accessible via an OAI-PMH handler or via a SPARQL endpoint.\n\nAs discussed above duplicates were removed from the dataset.", "#### Who are the source language producers?\n\nThe metadata producers are staff working in Italian cultural heritage institutions.", "### Annotations", "#### Annotation process\n\nFrom the paper:\n\n> \"Most of the dataset was manually annotated, with ~30K descriptions automatically labeled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections.\"\n\nTo determine the quality of the collected descriptions the authors of the paper used guidelines from the Istituto Centrale per il Catalogo e la Documentazione\n\nFrom the paper:\n\n> \"More precisely, a specific section of the guidelines addresses how to describe any cultural item, clarifying that both the object and the subject of the item must be presented in the description as follows:\n\n> Object: the object typology and shape must be described. To describe the object, the cataloguer must refer to the vocabularies provided by ICCD, using specific terminology (e.g. the technique used for paintings and drawings, or the material for the archaeological items);\n\n> Subject: the cataloguer must report the iconographic and decorative settings of the item, such as the characters of the depicted scene in a painting and their attribution. Other aspects (e.g. the history behind the painting or the painter) should not be included.\" p.221", "#### Who are the annotators?\n\n> \"The annotation is carried out by an expert in cultural heritage who collaborated in the past with Cultura Italia and has therefore in-depth knowledge of the data characteristics and of the ICCD guidelines.\" p.222", "### Personal and Sensitive Information\n\nNo personal or sensitive information is described in the paper.", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n- Lorenzini, Matteo\n- Rospocher, Marco\n- Tonelli, Sara", "### Licensing Information\n\ncc-by-4.0", "### Contributions\n\nThanks to @davanstrien for adding this dataset." ]

[ "TAGS\n#task_categories-text-classification #task_ids-acceptability-classification #annotations_creators-machine-generated #annotations_creators-expert-generated #language_creators-expert-generated #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-Italian #license-cc-by-4.0 #region-us \n", "# Dataset Card for Annotated dataset to assess the accuracy of the textual description of cultural heritage records", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage:URL\n- Repository:URL\n- Paper:URL\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nThe dataset contains more than 100K textual descriptions of cultural items from Cultura Italia, the Italian National Cultural aggregator. Each of the description is labeled either HIGH or LOW quality, according its adherence to the standard cataloguing guidelines provided by Istituto Centrale per il Catalogo e la Documentazione (ICCD). More precisely, each description is labeled as HIGH quality if the object and subject of the item (for which the description is provided) are both described according to the ICCD guidelines, and as LOW quality in all other cases. Most of the dataset was manually annotated, with ~30K descriptions automatically labeled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections. The dataset was developed to support the training and testing of ML text classification approaches for automatically assessing the quality of textual descriptions in digital Cultural Heritage repositories.", "### Supported Tasks and Leaderboards\n\nThis dataset can be used for text classification tasks. The paper introducing the dataset achieved an f1 score of '.783' for the task of classifying if a metadata record was low or high quality. Please see the results table for a full overview of the results reported in the paper.", "### Languages\n\nThe dataset consists of Italian metadata records. The labels are in English.", "## Dataset Structure\n\nThe dataset has only one configuration.", "### Data Instances\n\nAn example instance from the dataset:", "### Data Fields\n\nThe datafields are:\n\n- 'metadata_text': this contains the metadata text which was sourced from Cultura Italia\n- 'label': this is the label indicating if the record is 'High_Quality', or 'Low_Quality'. Most of the dataset was manually annotated, with ~30K descriptions automatically labelled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections.\n- 'source': the source of the metadata record", "### Data Splits\n\nThe dataset used 'ten-fold cross-validation' and doesn't report specific splits for train, validation and test data.", "## Dataset Creation\n\nThe dataset was generated using records from Cultura Italia. From the paper introducing the dataset:\n\n> By using the textual description encoded by the dc:description element from the Dublin Core metadata schema, we collect a dataset of 100,821 descriptions, after duplicate removal. These records include mainly data from “Musei d’Italia” and “Regione Marche” datasets, which have been chosen because they contain a high number of non-empty dc:description elements. p.221", "### Curation Rationale\n\nFrom the paper:\n\n> Duplicates were removed for two reasons: this reduced annotation effort in the subsequent manual annotation, and avoided that the same example appear both in the training and in the test set, a situation that could make classification biased and lead to inaccurate evaluation in supervised settings.Footnote 10 Duplicated descriptions were mainly short and of low-quality, reporting few generic words to describe an item (e.g. “Mensola.”, “Dipinto.”). p.221", "### Source Data", "#### Initial Data Collection and Normalization\n\nThe dataset was generated using records from Cultura Italia. This repository is accessible via an OAI-PMH handler or via a SPARQL endpoint.\n\nAs discussed above duplicates were removed from the dataset.", "#### Who are the source language producers?\n\nThe metadata producers are staff working in Italian cultural heritage institutions.", "### Annotations", "#### Annotation process\n\nFrom the paper:\n\n> \"Most of the dataset was manually annotated, with ~30K descriptions automatically labeled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections.\"\n\nTo determine the quality of the collected descriptions the authors of the paper used guidelines from the Istituto Centrale per il Catalogo e la Documentazione\n\nFrom the paper:\n\n> \"More precisely, a specific section of the guidelines addresses how to describe any cultural item, clarifying that both the object and the subject of the item must be presented in the description as follows:\n\n> Object: the object typology and shape must be described. To describe the object, the cataloguer must refer to the vocabularies provided by ICCD, using specific terminology (e.g. the technique used for paintings and drawings, or the material for the archaeological items);\n\n> Subject: the cataloguer must report the iconographic and decorative settings of the item, such as the characters of the depicted scene in a painting and their attribution. Other aspects (e.g. the history behind the painting or the painter) should not be included.\" p.221", "#### Who are the annotators?\n\n> \"The annotation is carried out by an expert in cultural heritage who collaborated in the past with Cultura Italia and has therefore in-depth knowledge of the data characteristics and of the ICCD guidelines.\" p.222", "### Personal and Sensitive Information\n\nNo personal or sensitive information is described in the paper.", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n- Lorenzini, Matteo\n- Rospocher, Marco\n- Tonelli, Sara", "### Licensing Information\n\ncc-by-4.0", "### Contributions\n\nThanks to @davanstrien for adding this dataset." ]

[ 107, 27, 125, 27, 219, 76, 22, 14, 14, 129, 36, 123, 124, 4, 59, 27, 5, 271, 57, 19, 8, 7, 8, 7, 5, 22, 12, 18 ]

[ "passage: TAGS\n#task_categories-text-classification #task_ids-acceptability-classification #annotations_creators-machine-generated #annotations_creators-expert-generated #language_creators-expert-generated #multilinguality-monolingual #size_categories-10K<n<100K #source_datasets-original #language-Italian #license-cc-by-4.0 #region-us \n# Dataset Card for Annotated dataset to assess the accuracy of the textual description of cultural heritage records## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage:URL\n- Repository:URL\n- Paper:URL\n- Leaderboard:\n- Point of Contact:### Dataset Summary\n\nThe dataset contains more than 100K textual descriptions of cultural items from Cultura Italia, the Italian National Cultural aggregator. Each of the description is labeled either HIGH or LOW quality, according its adherence to the standard cataloguing guidelines provided by Istituto Centrale per il Catalogo e la Documentazione (ICCD). More precisely, each description is labeled as HIGH quality if the object and subject of the item (for which the description is provided) are both described according to the ICCD guidelines, and as LOW quality in all other cases. Most of the dataset was manually annotated, with ~30K descriptions automatically labeled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections. The dataset was developed to support the training and testing of ML text classification approaches for automatically assessing the quality of textual descriptions in digital Cultural Heritage repositories.", "passage: ### Supported Tasks and Leaderboards\n\nThis dataset can be used for text classification tasks. The paper introducing the dataset achieved an f1 score of '.783' for the task of classifying if a metadata record was low or high quality. Please see the results table for a full overview of the results reported in the paper.### Languages\n\nThe dataset consists of Italian metadata records. The labels are in English.## Dataset Structure\n\nThe dataset has only one configuration.### Data Instances\n\nAn example instance from the dataset:### Data Fields\n\nThe datafields are:\n\n- 'metadata_text': this contains the metadata text which was sourced from Cultura Italia\n- 'label': this is the label indicating if the record is 'High_Quality', or 'Low_Quality'. Most of the dataset was manually annotated, with ~30K descriptions automatically labelled as LOW quality due to their length (less than 3 tokens) or their provenance from old (pre-2012), not curated, collections.\n- 'source': the source of the metadata record### Data Splits\n\nThe dataset used 'ten-fold cross-validation' and doesn't report specific splits for train, validation and test data.## Dataset Creation\n\nThe dataset was generated using records from Cultura Italia. From the paper introducing the dataset:\n\n> By using the textual description encoded by the dc:description element from the Dublin Core metadata schema, we collect a dataset of 100,821 descriptions, after duplicate removal. These records include mainly data from “Musei d’Italia” and “Regione Marche” datasets, which have been chosen because they contain a high number of non-empty dc:description elements. p.221### Curation Rationale\n\nFrom the paper:\n\n> Duplicates were removed for two reasons: this reduced annotation effort in the subsequent manual annotation, and avoided that the same example appear both in the training and in the test set, a situation that could make classification biased and lead to inaccurate evaluation in supervised settings.Footnote 10 Duplicated descriptions were mainly short and of low-quality, reporting few generic words to describe an item (e.g. “Mensola.”, “Dipinto.”). p.221### Source Data" ]

[ "annotations_creators:lexyr", "language_creators:crowdsourced", "multilinguality:monolingual", "size_categories:100K<n<1M", "source_datasets:original", "language:en", "license:cc-by-4.0", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for one-year-of-tsla-on-reddit", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Licensing Information", "## Dataset Description\n\n- Homepage: URL\n- Reddit downloader used: URL\n- Point of Contact: Website", "### Dataset Summary\n\nA year's worth of mentions of Tesla Inc. (TSLA) in Reddit posts and comments.", "### Languages\n\nMainly English.", "## Dataset Structure", "### Data Instances\n\nA data point is a post or a comment. Due to the separate nature of the two, those exist in two different files - even though many fields are shared.", "### Data Fields\n\n- 'type': the type of the data point. Can be 'post' or 'comment'.\n- 'id': the base-36 Reddit ID of the data point. Unique when combined with type.\n- 'URL': the base-36 Reddit ID of the data point's host subreddit. Unique.\n- 'URL': the human-readable name of the data point's host subreddit.\n- 'URL': a boolean marking the data point's host subreddit as NSFW or not.\n- 'created_utc': a UTC timestamp for the data point.\n- 'permalink': a reference link to the data point on Reddit.\n- 'score': score of the data point on Reddit.\n\n- 'domain': (Post only) the domain of the data point's link.\n- 'url': (Post only) the destination of the data point's link, if any.\n- 'selftext': (Post only) the self-text of the data point, if any.\n- 'title': (Post only) the title of the post data point.\n\n- 'body': (Comment only) the body of the comment data point.\n- 'sentiment': (Comment only) the result of an in-house sentiment analysis pipeline. Used for exploratory analysis.", "## Additional Information", "### Licensing Information\n\nCC-BY v4.0" ]

[ "TAGS\n#annotations_creators-lexyr #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-cc-by-4.0 #region-us \n", "# Dataset Card for one-year-of-tsla-on-reddit", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Licensing Information", "## Dataset Description\n\n- Homepage: URL\n- Reddit downloader used: URL\n- Point of Contact: Website", "### Dataset Summary\n\nA year's worth of mentions of Tesla Inc. (TSLA) in Reddit posts and comments.", "### Languages\n\nMainly English.", "## Dataset Structure", "### Data Instances\n\nA data point is a post or a comment. Due to the separate nature of the two, those exist in two different files - even though many fields are shared.", "### Data Fields\n\n- 'type': the type of the data point. Can be 'post' or 'comment'.\n- 'id': the base-36 Reddit ID of the data point. Unique when combined with type.\n- 'URL': the base-36 Reddit ID of the data point's host subreddit. Unique.\n- 'URL': the human-readable name of the data point's host subreddit.\n- 'URL': a boolean marking the data point's host subreddit as NSFW or not.\n- 'created_utc': a UTC timestamp for the data point.\n- 'permalink': a reference link to the data point on Reddit.\n- 'score': score of the data point on Reddit.\n\n- 'domain': (Post only) the domain of the data point's link.\n- 'url': (Post only) the destination of the data point's link, if any.\n- 'selftext': (Post only) the self-text of the data point, if any.\n- 'title': (Post only) the title of the post data point.\n\n- 'body': (Comment only) the body of the comment data point.\n- 'sentiment': (Comment only) the result of an in-house sentiment analysis pipeline. Used for exploratory analysis.", "## Additional Information", "### Licensing Information\n\nCC-BY v4.0" ]

[ 69, 18, 107, 21, 28, 8, 6, 41, 302, 5, 11 ]

[ "passage: TAGS\n#annotations_creators-lexyr #language_creators-crowdsourced #multilinguality-monolingual #size_categories-100K<n<1M #source_datasets-original #language-English #license-cc-by-4.0 #region-us \n# Dataset Card for one-year-of-tsla-on-reddit## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Licensing Information## Dataset Description\n\n- Homepage: URL\n- Reddit downloader used: URL\n- Point of Contact: Website### Dataset Summary\n\nA year's worth of mentions of Tesla Inc. (TSLA) in Reddit posts and comments.### Languages\n\nMainly English.## Dataset Structure### Data Instances\n\nA data point is a post or a comment. Due to the separate nature of the two, those exist in two different files - even though many fields are shared." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-large\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @abhijeet for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-large\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @abhijeet for evaluating this model." ]

[ 13, 70, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-large\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @abhijeet for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-base\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @abhijeet for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-base\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @abhijeet for evaluating this model." ]

[ 13, 69, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: t5-base\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @abhijeet for evaluating this model." ]

[ "license:other", "region:us" ]

[]

[]

[ "# Dataset Card for Stopword Lists for African Languages", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary", "### Context: \nSome words, like “the” or “and” in English, are used a lot in speech and writing. For most Natural Language Processing applications, you will want to remove these very frequent words. This is usually done using a list of “stopwords” which has been complied by hand.", "### Content: \nThis project uses the source texts provided by the African Storybook Project as a corpus and provides a number of tools to extract frequency lists and lists of stopwords from this corpus for the 60+ languages covered by ASP.\n\nIncluded in this dataset are the following languages:\n\n* Afrikaans: stoplist and word frequency\n* Hausa: stoplist and word frequency\n* Lugbarati: word frequency only\n* Lugbarati (Official): word frequency only\n* Somali: stoplist and word frequency\n* Sesotho: stoplist and word frequency\n* Kiswahili: stoplist and word frequency\n* Yoruba: stoplist and word frequency\n* isiZulu: stoplist and word frequency\n\nFiles are named using the language’s ISO code. For each language, URL is the list of stopwords, and code_frequency_list.txt is word frequency information. A list of ISO codes the the languages associated with them may be found in ISO_codes.csv.", "### Acknowledgements: \nThis project therefore attempts to fill in the gap in language coverage for African language stoplists by using the freely-available and open-licensed ASP Source project as a corpus.\nDual-licensed under CC-BY and Apache-2.0 license. Compiled by Liam Doherty. More information and the scripts used to generate these files are available here.", "### Inspiration: \nThis dataset is mainly helpful for use during NLP analysis, however there may some interesting insights in the data.\n\n* What qualities do stopwords share across languages? Given a novel language, could you predict what its stopwords should be?\n* What stopwords are shared across languages?\n* Often, related languages will have words with the same meaning and similar spellings. Can you automatically identify any of these pairs of words?", "### You may also like:\n\n* Stopword Lists for 19 Languages (mainly European and South Asian)", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis dataset was shared by @rtatman", "### Licensing Information\n\nThe license for this dataset is other", "### Contributions" ]

[ "TAGS\n#license-other #region-us \n", "# Dataset Card for Stopword Lists for African Languages", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary", "### Context: \nSome words, like “the” or “and” in English, are used a lot in speech and writing. For most Natural Language Processing applications, you will want to remove these very frequent words. This is usually done using a list of “stopwords” which has been complied by hand.", "### Content: \nThis project uses the source texts provided by the African Storybook Project as a corpus and provides a number of tools to extract frequency lists and lists of stopwords from this corpus for the 60+ languages covered by ASP.\n\nIncluded in this dataset are the following languages:\n\n* Afrikaans: stoplist and word frequency\n* Hausa: stoplist and word frequency\n* Lugbarati: word frequency only\n* Lugbarati (Official): word frequency only\n* Somali: stoplist and word frequency\n* Sesotho: stoplist and word frequency\n* Kiswahili: stoplist and word frequency\n* Yoruba: stoplist and word frequency\n* isiZulu: stoplist and word frequency\n\nFiles are named using the language’s ISO code. For each language, URL is the list of stopwords, and code_frequency_list.txt is word frequency information. A list of ISO codes the the languages associated with them may be found in ISO_codes.csv.", "### Acknowledgements: \nThis project therefore attempts to fill in the gap in language coverage for African language stoplists by using the freely-available and open-licensed ASP Source project as a corpus.\nDual-licensed under CC-BY and Apache-2.0 license. Compiled by Liam Doherty. More information and the scripts used to generate these files are available here.", "### Inspiration: \nThis dataset is mainly helpful for use during NLP analysis, however there may some interesting insights in the data.\n\n* What qualities do stopwords share across languages? Given a novel language, could you predict what its stopwords should be?\n* What stopwords are shared across languages?\n* Often, related languages will have words with the same meaning and similar spellings. Can you automatically identify any of these pairs of words?", "### You may also like:\n\n* Stopword Lists for 19 Languages (mainly European and South Asian)", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis dataset was shared by @rtatman", "### Licensing Information\n\nThe license for this dataset is other", "### Contributions" ]

[ 11, 13, 125, 25, 6, 67, 233, 88, 101, 24, 10, 4, 6, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 16, 14, 5 ]

[ "passage: TAGS\n#license-other #region-us \n# Dataset Card for Stopword Lists for African Languages## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:### Dataset Summary### Context: \nSome words, like “the” or “and” in English, are used a lot in speech and writing. For most Natural Language Processing applications, you will want to remove these very frequent words. This is usually done using a list of “stopwords” which has been complied by hand.### Content: \nThis project uses the source texts provided by the African Storybook Project as a corpus and provides a number of tools to extract frequency lists and lists of stopwords from this corpus for the 60+ languages covered by ASP.\n\nIncluded in this dataset are the following languages:\n\n* Afrikaans: stoplist and word frequency\n* Hausa: stoplist and word frequency\n* Lugbarati: word frequency only\n* Lugbarati (Official): word frequency only\n* Somali: stoplist and word frequency\n* Sesotho: stoplist and word frequency\n* Kiswahili: stoplist and word frequency\n* Yoruba: stoplist and word frequency\n* isiZulu: stoplist and word frequency\n\nFiles are named using the language’s ISO code. For each language, URL is the list of stopwords, and code_frequency_list.txt is word frequency information. A list of ISO codes the the languages associated with them may be found in ISO_codes.csv." ]

[ "license:cc0-1.0", "region:us" ]

[]

[]

[ "# Dataset Card for Airbnb Stock Price", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nThis contains the historical stock price of Airbnb (ticker symbol ABNB) an American company that operates an online marketplace for lodging, primarily homestays for vacation rentals, and tourism activities. Based in San Francisco, California, the platform is accessible via website and mobile app.", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis dataset was shared by @evangower", "### Licensing Information\n\nThe license for this dataset is cc0-1.0", "### Contributions" ]

[ "TAGS\n#license-cc0-1.0 #region-us \n", "# Dataset Card for Airbnb Stock Price", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nThis contains the historical stock price of Airbnb (ticker symbol ABNB) an American company that operates an online marketplace for lodging, primarily homestays for vacation rentals, and tourism activities. Based in San Francisco, California, the platform is accessible via website and mobile app.", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis dataset was shared by @evangower", "### Licensing Information\n\nThe license for this dataset is cc0-1.0", "### Contributions" ]

[ 14, 8, 125, 25, 68, 10, 4, 6, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 16, 18, 5 ]

[ "passage: TAGS\n#license-cc0-1.0 #region-us \n# Dataset Card for Airbnb Stock Price## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:### Dataset Summary\n\nThis contains the historical stock price of Airbnb (ticker symbol ABNB) an American company that operates an online marketplace for lodging, primarily homestays for vacation rentals, and tourism activities. Based in San Francisco, California, the platform is accessible via website and mobile app.### Supported Tasks and Leaderboards### Languages## Dataset Structure### Data Instances### Data Fields### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data### Social Impact of Dataset### Discussion of Biases### Other Known Limitations## Additional Information### Dataset Curators\n\nThis dataset was shared by @evangower### Licensing Information\n\nThe license for this dataset is cc0-1.0### Contributions" ]

[ "license:other", "region:us" ]

[]

[]

[ "# Dataset Card for Pizza or Not Pizza?", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nWho doesn't like pizza? This dataset contains about 1000 images of pizza and 1000 images of dishes other than pizza. It can be used for a simple binary image classification task.\n\nAll images were rescaled to have a maximum side length of 512 pixels. \n\nThis is a subset of the Food-101 dataset. Information about the original dataset can be found in the following paper: \nBossard, Lukas, Matthieu Guillaumin, and Luc Van Gool. \"Food-101 – Mining Discriminative Components with Random Forests.\" In *European conference on computer vision*, pp. 446-461. Springer, Cham, 2014.\n\nThe original dataset can be found in the following locations:\nURL\nURL\nURL\nURL\n\nNumber of instances in each class:\nPizza: 983\nNot Pizza: 983", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis dataset was shared by @carlosrunner", "### Licensing Information\n\nThe license for this dataset is other", "### Contributions" ]

[ "TAGS\n#license-other #region-us \n", "# Dataset Card for Pizza or Not Pizza?", "## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:", "### Dataset Summary\n\nWho doesn't like pizza? This dataset contains about 1000 images of pizza and 1000 images of dishes other than pizza. It can be used for a simple binary image classification task.\n\nAll images were rescaled to have a maximum side length of 512 pixels. \n\nThis is a subset of the Food-101 dataset. Information about the original dataset can be found in the following paper: \nBossard, Lukas, Matthieu Guillaumin, and Luc Van Gool. \"Food-101 – Mining Discriminative Components with Random Forests.\" In *European conference on computer vision*, pp. 446-461. Springer, Cham, 2014.\n\nThe original dataset can be found in the following locations:\nURL\nURL\nURL\nURL\n\nNumber of instances in each class:\nPizza: 983\nNot Pizza: 983", "### Supported Tasks and Leaderboards", "### Languages", "## Dataset Structure", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators\n\nThis dataset was shared by @carlosrunner", "### Licensing Information\n\nThe license for this dataset is other", "### Contributions" ]

[ 11, 10, 125, 25, 183, 10, 4, 6, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 16, 14, 5 ]

[ "passage: TAGS\n#license-other #region-us \n# Dataset Card for Pizza or Not Pizza?## Table of Contents\n- Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository:\n- Paper:\n- Leaderboard:\n- Point of Contact:### Dataset Summary\n\nWho doesn't like pizza? This dataset contains about 1000 images of pizza and 1000 images of dishes other than pizza. It can be used for a simple binary image classification task.\n\nAll images were rescaled to have a maximum side length of 512 pixels. \n\nThis is a subset of the Food-101 dataset. Information about the original dataset can be found in the following paper: \nBossard, Lukas, Matthieu Guillaumin, and Luc Van Gool. \"Food-101 – Mining Discriminative Components with Random Forests.\" In *European conference on computer vision*, pp. 446-461. Springer, Cham, 2014.\n\nThe original dataset can be found in the following locations:\nURL\nURL\nURL\nURL\n\nNumber of instances in each class:\nPizza: 983\nNot Pizza: 983### Supported Tasks and Leaderboards### Languages## Dataset Structure### Data Instances### Data Fields### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data### Social Impact of Dataset### Discussion of Biases### Other Known Limitations## Additional Information### Dataset Curators\n\nThis dataset was shared by @carlosrunner" ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"simple-wiki\"", "## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: David Kauchak", "### Dataset Summary\nThis dataset contains pairs of equivalent sentences obtained from Wikipedia.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n", "# Dataset Card for \"simple-wiki\"", "## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: David Kauchak", "### Dataset Summary\nThis dataset contains pairs of equivalent sentences obtained from Wikipedia.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ 43, 10, 120, 26, 22, 24, 7, 76, 59, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 5 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n# Dataset Card for \"simple-wiki\"## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: David Kauchak### Dataset Summary\nThis dataset contains pairs of equivalent sentences obtained from Wikipedia.### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n- English.## Dataset Structure\nEach example in the dataset contains pairs of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data### Social Impact of Dataset### Discussion of Biases### Other Known Limitations## Additional Information### Dataset Curators### Licensing Information### Contributions" ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"sentence-compression\"", "## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Katja Filippova\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 14.2 MB", "### Dataset Summary\nDataset with pairs of equivalent sentences.\nThe dataset is provided \"AS IS\" without any warranty, express or implied. \nGoogle disclaims all liability for any damages, direct or indirect, resulting from using the dataset.\n\nDisclaimer: The team releasing sentence-compression did not upload the dataset to the Hub and did not write a dataset card. These steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n", "# Dataset Card for \"sentence-compression\"", "## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Katja Filippova\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 14.2 MB", "### Dataset Summary\nDataset with pairs of equivalent sentences.\nThe dataset is provided \"AS IS\" without any warranty, express or implied. \nGoogle disclaims all liability for any damages, direct or indirect, resulting from using the dataset.\n\nDisclaimer: The team releasing sentence-compression did not upload the dataset to the Hub and did not write a dataset card. These steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ 43, 12, 120, 53, 99, 24, 7, 79, 59, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 5 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n# Dataset Card for \"sentence-compression\"## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Katja Filippova\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 14.2 MB### Dataset Summary\nDataset with pairs of equivalent sentences.\nThe dataset is provided \"AS IS\" without any warranty, express or implied. \nGoogle disclaims all liability for any damages, direct or indirect, resulting from using the dataset.\n\nDisclaimer: The team releasing sentence-compression did not upload the dataset to the Hub and did not write a dataset card. These steps were done by the Hugging Face team.### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n- English.## Dataset Structure\nEach example in the dataset contains pairs of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:### Curation Rationale### Source Data" ]

[ "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"altlex\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: Christopher Hidey", "### Dataset Summary\n\nGit repository for software associated with the 2016 ACL paper \"Identifying Causal Relations Using Parallel Wikipedia Articles.\"\n\nDisclaimer: The team releasing altlex did not upload the dataset to the Hub and did not write a dataset card.\nThese steps were done by the Hugging Face team.", "### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n\n- English.", "## Dataset Structure\n\nEach example in the dataset contains a pair of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\n\n\nReview an example 'i' with:", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions\n\n- @chridey for adding this dataset to Github.\n\n---" ]

[ "TAGS\n#language-English #license-mit #region-us \n", "# Dataset Card for \"altlex\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: Christopher Hidey", "### Dataset Summary\n\nGit repository for software associated with the 2016 ACL paper \"Identifying Causal Relations Using Parallel Wikipedia Articles.\"\n\nDisclaimer: The team releasing altlex did not upload the dataset to the Hub and did not write a dataset card.\nThese steps were done by the Hugging Face team.", "### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n\n- English.", "## Dataset Structure\n\nEach example in the dataset contains a pair of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\n\n\nReview an example 'i' with:", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions\n\n- @chridey for adding this dataset to Github.\n\n---" ]

[ 15, 9, 120, 25, 73, 24, 7, 80, 59, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 21 ]

[ "passage: TAGS\n#language-English #license-mit #region-us \n# Dataset Card for \"altlex\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: Christopher Hidey### Dataset Summary\n\nGit repository for software associated with the 2016 ACL paper \"Identifying Causal Relations Using Parallel Wikipedia Articles.\"\n\nDisclaimer: The team releasing altlex did not upload the dataset to the Hub and did not write a dataset card.\nThese steps were done by the Hugging Face team.### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n\n- English.## Dataset Structure\n\nEach example in the dataset contains a pair of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\n\n\nReview an example 'i' with:### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?### Personal and Sensitive Information## Considerations for Using the Data### Social Impact of Dataset### Discussion of Biases### Other Known Limitations## Additional Information### Dataset Curators" ]

[ "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"flickr30k-captions\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Usage Example\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: Peter Young, Alice Lai, Micah Hodosh, Julia Hockenmaier", "### Dataset Summary\n\nWe propose to use the visual denotations of linguistic expressions (i.e. the set of images they describe) to define novel denotational similarity metrics, which we show to be at least as beneficial as distributional similarities for two tasks that require semantic inference. To compute these denotational similarities, we construct a denotation graph, i.e. a subsumption hierarchy over constituents and their denotations, based on a large corpus of 30K images and 150K descriptive captions.\n\nDisclaimer: The team releasing Flickr30k did not upload the dataset to the Hub and did not write a dataset card. These steps were done by the Hugging Face team.", "### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n\n- English.", "## Dataset Structure\n\nEach example in the dataset contains quintets of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' has the format:\n\n\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions\n\nThanks to Peter Young, Alice Lai, Micah Hodosh, Julia Hockenmaier for adding this dataset." ]

[ "TAGS\n#language-English #license-mit #region-us \n", "# Dataset Card for \"flickr30k-captions\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Usage Example\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: Peter Young, Alice Lai, Micah Hodosh, Julia Hockenmaier", "### Dataset Summary\n\nWe propose to use the visual denotations of linguistic expressions (i.e. the set of images they describe) to define novel denotational similarity metrics, which we show to be at least as beneficial as distributional similarities for two tasks that require semantic inference. To compute these denotational similarities, we construct a denotation graph, i.e. a subsumption hierarchy over constituents and their denotations, based on a large corpus of 30K images and 150K descriptive captions.\n\nDisclaimer: The team releasing Flickr30k did not upload the dataset to the Hub and did not write a dataset card. These steps were done by the Hugging Face team.", "### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n\n- English.", "## Dataset Structure\n\nEach example in the dataset contains quintets of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' has the format:\n\n\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions\n\nThanks to Peter Young, Alice Lai, Micah Hodosh, Julia Hockenmaier for adding this dataset." ]

[ 15, 15, 125, 38, 165, 24, 7, 81, 58, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 29 ]

[ "passage: TAGS\n#language-English #license-mit #region-us \n# Dataset Card for \"flickr30k-captions\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Usage Example\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: URL\n- Point of Contact: Peter Young, Alice Lai, Micah Hodosh, Julia Hockenmaier### Dataset Summary\n\nWe propose to use the visual denotations of linguistic expressions (i.e. the set of images they describe) to define novel denotational similarity metrics, which we show to be at least as beneficial as distributional similarities for two tasks that require semantic inference. To compute these denotational similarities, we construct a denotation graph, i.e. a subsumption hierarchy over constituents and their denotations, based on a large corpus of 30K images and 150K descriptive captions.\n\nDisclaimer: The team releasing Flickr30k did not upload the dataset to the Hub and did not write a dataset card. These steps were done by the Hugging Face team.### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n\n- English.## Dataset Structure\n\nEach example in the dataset contains quintets of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences." ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "arxiv:1405.0312", "region:us" ]

[ "1405.0312" ]

[ "en" ]

[ "# Dataset Card for \"coco_captions\"", "## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper: \n- Point of Contact: info@URL\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 6.32 MB", "### Dataset Summary\n\nCOCO is a large-scale object detection, segmentation, and captioning dataset. This repo contains five captions per image; useful for sentence similarity tasks.\n\nDisclaimer: The team releasing COCO did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n\n- English.", "## Dataset Structure\n\nEach example in the dataset contains quintets of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\n\n\nReview an example 'i' with:", "### Data Instances", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\nThe annotations in this dataset along with this website belong to the COCO Consortium \nand are licensed under a Creative Commons Attribution 4.0 License", "### Contributions\n\nThanks to:\n\n- Tsung-Yi Lin - Google Brain\n- Genevieve Patterson - MSR, Trash TV\n- Matteo R. - Ronchi Caltech\n- Yin Cui - Google\n- Michael Maire - TTI-Chicago\n- Serge Belongie - Cornell Tech\n- Lubomir Bourdev - WaveOne, Inc.\n- Ross Girshick - FAIR\n- James Hays - Georgia Tech\n- Pietro Perona - Caltech\n- Deva Ramanan - CMU\n- Larry Zitnick - FAIR\n- Piotr Dollár - FAIR\n\nfor adding this dataset." ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #arxiv-1405.0312 #region-us \n", "# Dataset Card for \"coco_captions\"", "## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper: \n- Point of Contact: info@URL\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 6.32 MB", "### Dataset Summary\n\nCOCO is a large-scale object detection, segmentation, and captioning dataset. This repo contains five captions per image; useful for sentence similarity tasks.\n\nDisclaimer: The team releasing COCO did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n\n- English.", "## Dataset Structure\n\nEach example in the dataset contains quintets of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\n\n\nReview an example 'i' with:", "### Data Instances", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\nThe annotations in this dataset along with this website belong to the COCO Consortium \nand are licensed under a Creative Commons Attribution 4.0 License", "### Contributions\n\nThanks to:\n\n- Tsung-Yi Lin - Google Brain\n- Genevieve Patterson - MSR, Trash TV\n- Matteo R. - Ronchi Caltech\n- Yin Cui - Google\n- Michael Maire - TTI-Chicago\n- Serge Belongie - Cornell Tech\n- Lubomir Bourdev - WaveOne, Inc.\n- Ross Girshick - FAIR\n- James Hays - Georgia Tech\n- Pietro Perona - Caltech\n- Deva Ramanan - CMU\n- Larry Zitnick - FAIR\n- Piotr Dollár - FAIR\n\nfor adding this dataset." ]

[ 52, 12, 120, 53, 83, 24, 7, 81, 59, 6, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 38, 131 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #arxiv-1405.0312 #region-us \n# Dataset Card for \"coco_captions\"## Table of Contents\n\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n- Homepage: URL\n- Repository: URL\n- Paper: \n- Point of Contact: info@URL\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 6.32 MB### Dataset Summary\n\nCOCO is a large-scale object detection, segmentation, and captioning dataset. This repo contains five captions per image; useful for sentence similarity tasks.\n\nDisclaimer: The team releasing COCO did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.### Supported Tasks\n\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n\n- English.## Dataset Structure\n\nEach example in the dataset contains quintets of similar sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\":\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\n\n\nReview an example 'i' with:### Data Instances### Data Splits## Dataset Creation" ]

[ "task_categories:text-generation", "task_ids:language-modeling", "multilinguality:other-music", "license:cc", "arxiv:2008.06048", "region:us" ]

[ "2008.06048" ]

[]

[ "# Mutopia Guitar Dataset", "## Table of Contents\n- Dataset Card Creation Guide\n - Table of Contents\n - Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits", "## Dataset Description\n\n- Homepage: Mutopia Project\n- Repository implementation of the paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer and the Johann Sebastian Bach Chorales Dataset\n- Based on Paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer\n- Point of Contact: Juan Carlos Piñeros", "### Dataset Summary\n\nMutopia guitar dataset consists of the soloist guitar pieces of the Mutopia Project. I encoded the MIDI files into text tokens using the excellent implementation of Dr. Tristan Beheren of the paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer. \n\nThe dataset mainly contains guitar music from western classical composers, such as Sor, Aguado, Carcassi, and Giuliani.", "### Supported Tasks and Leaderboards\n\nAnyone interested can use the dataset to train a model for symbolic music generation, which consists in treating symbols for music sounds (notes) as text tokens. Then, one can implement a generative model using NLP techniques, such as Transformers.", "## Dataset Structure", "### Data Instances\n\nEach guitar piece is represented as a line of text that contains a series of tokens, for instance:\n\nPIECE_START: Where the piece begins\nPIECE_ENDS: Where the piece ends\nTIME_SIGNATURE: Time signature for the piece\nBPM: Tempo of the piece\nBAR_START: Begining of a new bar\nNOTE_ON: Start of a new musical note specifying its MIDI note number\nTIME_DELTA: Duration until the next event\nNOTE_OFF: End of musical note specifying its MIDI note number", "### Data Fields\n\n- 'text': Sequence of tokens that represent the guitar piece as explained in the paper MMM: Exploring Conditional Multi-Track Music Generation with the Transformer.", "### Data Splits\n\nThere are, at this moment, 395 MIDI guitar files in the Mutopia Project. I removed files of pieces that were not music for soloist guitar. After this removal, there were 372 MIDI files.\n\nI used an 80/20 split and augmented the training dataset by transposing the piece 1 octave above and below (24 semitones). The final result is then:\n\nTrain dataset: 7325 pieces\nTest dataset: 74 pieces" ]

[ "TAGS\n#task_categories-text-generation #task_ids-language-modeling #multilinguality-other-music #license-cc #arxiv-2008.06048 #region-us \n", "# Mutopia Guitar Dataset", "## Table of Contents\n- Dataset Card Creation Guide\n - Table of Contents\n - Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits", "## Dataset Description\n\n- Homepage: Mutopia Project\n- Repository implementation of the paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer and the Johann Sebastian Bach Chorales Dataset\n- Based on Paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer\n- Point of Contact: Juan Carlos Piñeros", "### Dataset Summary\n\nMutopia guitar dataset consists of the soloist guitar pieces of the Mutopia Project. I encoded the MIDI files into text tokens using the excellent implementation of Dr. Tristan Beheren of the paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer. \n\nThe dataset mainly contains guitar music from western classical composers, such as Sor, Aguado, Carcassi, and Giuliani.", "### Supported Tasks and Leaderboards\n\nAnyone interested can use the dataset to train a model for symbolic music generation, which consists in treating symbols for music sounds (notes) as text tokens. Then, one can implement a generative model using NLP techniques, such as Transformers.", "## Dataset Structure", "### Data Instances\n\nEach guitar piece is represented as a line of text that contains a series of tokens, for instance:\n\nPIECE_START: Where the piece begins\nPIECE_ENDS: Where the piece ends\nTIME_SIGNATURE: Time signature for the piece\nBPM: Tempo of the piece\nBAR_START: Begining of a new bar\nNOTE_ON: Start of a new musical note specifying its MIDI note number\nTIME_DELTA: Duration until the next event\nNOTE_OFF: End of musical note specifying its MIDI note number", "### Data Fields\n\n- 'text': Sequence of tokens that represent the guitar piece as explained in the paper MMM: Exploring Conditional Multi-Track Music Generation with the Transformer.", "### Data Splits\n\nThere are, at this moment, 395 MIDI guitar files in the Mutopia Project. I removed files of pieces that were not music for soloist guitar. After this removal, there were 372 MIDI files.\n\nI used an 80/20 split and augmented the training dataset by transposing the piece 1 octave above and below (24 semitones). The final result is then:\n\nTrain dataset: 7325 pieces\nTest dataset: 74 pieces" ]

[ 48, 6, 54, 78, 102, 68, 6, 123, 44, 101 ]

[ "passage: TAGS\n#task_categories-text-generation #task_ids-language-modeling #multilinguality-other-music #license-cc #arxiv-2008.06048 #region-us \n# Mutopia Guitar Dataset## Table of Contents\n- Dataset Card Creation Guide\n - Table of Contents\n - Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits## Dataset Description\n\n- Homepage: Mutopia Project\n- Repository implementation of the paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer and the Johann Sebastian Bach Chorales Dataset\n- Based on Paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer\n- Point of Contact: Juan Carlos Piñeros### Dataset Summary\n\nMutopia guitar dataset consists of the soloist guitar pieces of the Mutopia Project. I encoded the MIDI files into text tokens using the excellent implementation of Dr. Tristan Beheren of the paper: MMM: Exploring Conditional Multi-Track Music Generation with the Transformer. \n\nThe dataset mainly contains guitar music from western classical composers, such as Sor, Aguado, Carcassi, and Giuliani.### Supported Tasks and Leaderboards\n\nAnyone interested can use the dataset to train a model for symbolic music generation, which consists in treating symbols for music sounds (notes) as text tokens. Then, one can implement a generative model using NLP techniques, such as Transformers.## Dataset Structure### Data Instances\n\nEach guitar piece is represented as a line of text that contains a series of tokens, for instance:\n\nPIECE_START: Where the piece begins\nPIECE_ENDS: Where the piece ends\nTIME_SIGNATURE: Time signature for the piece\nBPM: Tempo of the piece\nBAR_START: Begining of a new bar\nNOTE_ON: Start of a new musical note specifying its MIDI note number\nTIME_DELTA: Duration until the next event\nNOTE_OFF: End of musical note specifying its MIDI note number" ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "arxiv:2004.07180", "region:us" ]

[ "2004.07180" ]

[ "en" ]

[ "# Dataset Card for \"SPECTER\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: @armancohan, @sergeyf, @haroldrubio, @jinamshah", "### Dataset Summary\n\nDataset containing triplets (three sentences): anchor, positive, and negative. Contains titles of papers.\n\nDisclaimer: The team releasing SPECTER did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "## Dataset Structure\nEach example in the dataset contains triplets of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nEach example is a dictionary with a key, \"set\", containing a list of three sentences (anchor, positive, and negative):\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using triplets.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #arxiv-2004.07180 #region-us \n", "# Dataset Card for \"SPECTER\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: @armancohan, @sergeyf, @haroldrubio, @jinamshah", "### Dataset Summary\n\nDataset containing triplets (three sentences): anchor, positive, and negative. Contains titles of papers.\n\nDisclaimer: The team releasing SPECTER did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "## Dataset Structure\nEach example in the dataset contains triplets of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nEach example is a dictionary with a key, \"set\", containing a list of three sentences (anchor, positive, and negative):\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using triplets.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ 52, 9, 120, 42, 74, 106, 59, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 5 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #arxiv-2004.07180 #region-us \n# Dataset Card for \"SPECTER\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: @armancohan, @sergeyf, @haroldrubio, @jinamshah### Dataset Summary\n\nDataset containing triplets (three sentences): anchor, positive, and negative. Contains titles of papers.\n\nDisclaimer: The team releasing SPECTER did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.## Dataset Structure\nEach example in the dataset contains triplets of equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nEach example is a dictionary with a key, \"set\", containing a list of three sentences (anchor, positive, and negative):\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using triplets.### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process" ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: Ameer05/bart-large-cnn-samsum-rescom-finetuned-resume-summarizer-10-epoch\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @Abhijeet3922 for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: Ameer05/bart-large-cnn-samsum-rescom-finetuned-resume-summarizer-10-epoch\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @Abhijeet3922 for evaluating this model." ]

[ 13, 97, 19 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: Ameer05/bart-large-cnn-samsum-rescom-finetuned-resume-summarizer-10-epoch\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @Abhijeet3922 for evaluating this model." ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"QQP_triplets\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Kornél Csernai, Nikhil Dandekar, Shankar Iyer", "### Dataset Summary\n\nThis dataset will give anyone the opportunity to train and test models of semantic equivalence, based on actual Quora data. The data is organized as triplets (anchor, positive, negative).\n\nDisclaimer: The team releasing Quora data did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example is a dictionary with three keys (query, pos, and neg) containing a list each (triplets). The first key contains an anchor sentence, the second a positive sentence, and the third a list of negative sentences. \n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train them.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\nHere are a few important things to keep in mind about this dataset:\n\n- Our original sampling method returned an imbalanced dataset with many more true examples of duplicate pairs than non-duplicates. \nTherefore, we supplemented the dataset with negative examples. \n- One source of negative examples were pairs of “related questions” which, although pertaining to similar topics, \nare not truly semantically equivalent.\n- The distribution of questions in the dataset should not be taken to be representative of the distribution of questions asked on Quora. This is, in part, because of the combination of sampling procedures and also due to some sanitization measures that\nhave been applied to the final dataset (e.g., removal of questions with extremely long question details).\n- The ground-truth labels contain some amount of noise: they are not guaranteed to be perfect.", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions\n\nThanks to Kornél Csernai, Nikhil Dandekar, Shankar Iyer for adding this dataset." ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n", "# Dataset Card for \"QQP_triplets\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Kornél Csernai, Nikhil Dandekar, Shankar Iyer", "### Dataset Summary\n\nThis dataset will give anyone the opportunity to train and test models of semantic equivalence, based on actual Quora data. The data is organized as triplets (anchor, positive, negative).\n\nDisclaimer: The team releasing Quora data did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example is a dictionary with three keys (query, pos, and neg) containing a list each (triplets). The first key contains an anchor sentence, the second a positive sentence, and the third a list of negative sentences. \n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train them.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations\n\nHere are a few important things to keep in mind about this dataset:\n\n- Our original sampling method returned an imbalanced dataset with many more true examples of duplicate pairs than non-duplicates. \nTherefore, we supplemented the dataset with negative examples. \n- One source of negative examples were pairs of “related questions” which, although pertaining to similar topics, \nare not truly semantically equivalent.\n- The distribution of questions in the dataset should not be taken to be representative of the distribution of questions asked on Quora. This is, in part, because of the combination of sampling procedures and also due to some sanitization measures that\nhave been applied to the final dataset (e.g., removal of questions with extremely long question details).\n- The ground-truth labels contain some amount of noise: they are not guaranteed to be perfect.", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions\n\nThanks to Kornél Csernai, Nikhil Dandekar, Shankar Iyer for adding this dataset." ]

[ 43, 13, 120, 37, 87, 24, 7, 88, 59, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 202, 5, 6, 6, 29 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n# Dataset Card for \"QQP_triplets\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Kornél Csernai, Nikhil Dandekar, Shankar Iyer### Dataset Summary\n\nThis dataset will give anyone the opportunity to train and test models of semantic equivalence, based on actual Quora data. The data is organized as triplets (anchor, positive, negative).\n\nDisclaimer: The team releasing Quora data did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n- English.## Dataset Structure\nEach example is a dictionary with three keys (query, pos, and neg) containing a list each (triplets). The first key contains an anchor sentence, the second a positive sentence, and the third a list of negative sentences. \n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train them.### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:### Curation Rationale### Source Data#### Initial Data Collection and Normalization" ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-large\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-large\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 74, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-large\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-xsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-xsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 74, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-xsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-cnn_dailymail\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-cnn_dailymail\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 77, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-cnn_dailymail\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-reddit_tifu\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-reddit_tifu\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 77, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-reddit_tifu\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: facebook/bart-large-xsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: facebook/bart-large-xsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 75, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: facebook/bart-large-xsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: knkarthick/bart-large-xsum-samsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: knkarthick/bart-large-xsum-samsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 81, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: knkarthick/bart-large-xsum-samsum\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: facebook/bart-large-cnn\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: facebook/bart-large-cnn\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ 13, 75, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: facebook/bart-large-cnn\n* Dataset: samsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @ikadebi for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-xsum\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @sheikmohdimran for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-xsum\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @sheikmohdimran for evaluating this model." ]

[ 13, 74, 19 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-xsum\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @sheikmohdimran for evaluating this model." ]

[ "autotrain", "evaluation", "region:us" ]

[]

[]

[ "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-wikihow\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @Scofield for evaluating this model." ]

[ "TAGS\n#autotrain #evaluation #region-us \n", "# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-wikihow\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.", "## Contributions\n\nThanks to @Scofield for evaluating this model." ]

[ 13, 74, 16 ]

[ "passage: TAGS\n#autotrain #evaluation #region-us \n# Dataset Card for AutoTrain Evaluator\n\nThis repository contains model predictions generated by AutoTrain for the following task and dataset:\n\n* Task: Summarization\n* Model: google/pegasus-wikihow\n* Dataset: xsum\n\nTo run new evaluation jobs, visit Hugging Face's automatic model evaluator.## Contributions\n\nThanks to @Scofield for evaluating this model." ]

[ "benchmark:gem", "evaluation", "benchmark", "region:us" ]

[]

[]

[ "# GEM Submission\n\nSubmission name: This is a test submission 3" ]

[ "TAGS\n#benchmark-gem #evaluation #benchmark #region-us \n", "# GEM Submission\n\nSubmission name: This is a test submission 3" ]

[ 19, 16 ]

[ "passage: TAGS\n#benchmark-gem #evaluation #benchmark #region-us \n# GEM Submission\n\nSubmission name: This is a test submission 3" ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-unique\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-unique\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 28, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-unique\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"Amazon-QA\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Julian McAuley\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 247 MB", "### Dataset Summary\n\nThis dataset contains Question and Answer data from Amazon.\n\nDisclaimer: The team releasing Amazon-QA did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of query and answer sentences and is formatted as a dictionary:\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n", "# Dataset Card for \"Amazon-QA\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Julian McAuley\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 247 MB", "### Dataset Summary\n\nThis dataset contains Question and Answer data from Amazon.\n\nDisclaimer: The team releasing Amazon-QA did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of query and answer sentences and is formatted as a dictionary:\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ 43, 10, 120, 52, 57, 24, 7, 62, 59, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 5 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n# Dataset Card for \"Amazon-QA\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Julian McAuley\n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 247 MB### Dataset Summary\n\nThis dataset contains Question and Answer data from Amazon.\n\nDisclaimer: The team releasing Amazon-QA did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n- English.## Dataset Structure\nEach example in the dataset contains pairs of query and answer sentences and is formatted as a dictionary:\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:### Data Instances### Data Fields### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization#### Who are the source language producers?### Annotations#### Annotation process#### Who are the annotators?" ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "arxiv:2102.07033", "region:us" ]

[ "2102.07033" ]

[ "en" ]

[ "# Dataset Card for \"PAQ_pairs\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: \n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 21 Bytes", "### Dataset Summary\n\nPairs questions and answers obtained from Wikipedia.\n\nDisclaimer: The team releasing PAQ QA pairs did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\". The first sentence is a question and the second an answer; thus, both sentences would be similar.\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\nThe PAQ QA-pairs and metadata is licensed under CC-BY-SA. \nOther data is licensed according to the accompanying license files.", "### Contributions\n\nThanks to @patrick-s-h-lewis for adding this dataset." ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #arxiv-2102.07033 #region-us \n", "# Dataset Card for \"PAQ_pairs\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: \n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 21 Bytes", "### Dataset Summary\n\nPairs questions and answers obtained from Wikipedia.\n\nDisclaimer: The team releasing PAQ QA pairs did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains pairs of sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\". The first sentence is a question and the second an answer; thus, both sentences would be similar.\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.", "### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information\n\nThe PAQ QA-pairs and metadata is licensed under CC-BY-SA. \nOther data is licensed according to the accompanying license files.", "### Contributions\n\nThanks to @patrick-s-h-lewis for adding this dataset." ]

[ 51, 12, 120, 49, 59, 24, 7, 100, 59, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 41, 23 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #arxiv-2102.07033 #region-us \n# Dataset Card for \"PAQ_pairs\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: \n- Size of downloaded dataset files: \n- Size of the generated dataset: \n- Total amount of disk used: 21 Bytes### Dataset Summary\n\nPairs questions and answers obtained from Wikipedia.\n\nDisclaimer: The team releasing PAQ QA pairs did not upload the dataset to the Hub and did not write a dataset card. \nThese steps were done by the Hugging Face team.### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n- English.## Dataset Structure\nEach example in the dataset contains pairs of sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\". The first sentence is a question and the second an answer; thus, both sentences would be similar.\n\n\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar pairs of sentences.### Usage Example\n\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format:\n\nReview an example 'i' with:### Data Instances### Data Fields### Data Splits## Dataset Creation" ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster-noise\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster-noise\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 31, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster-noise\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster00\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster00\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster00\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster01\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster01\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster01\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster10\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster10\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster10\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster11\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster11\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster11\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster12\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster12\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster12\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster13\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster13\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster13\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster14\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster14\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster14\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster15\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster15\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster15\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster16\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster16\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster16\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster17\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster17\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster17\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster18\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster18\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster18\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster19\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster19\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster19\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster02\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster02\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster02\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster20\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster20\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster20\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster21\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster21\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster21\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster22\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster22\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster22\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster23\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster23\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster23\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster24\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster24\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster24\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster25\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster25\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster25\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster26\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster26\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster26\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster27\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster27\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster27\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster28\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster28\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster28\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster29\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster29\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster29\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster03\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster03\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster03\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster04\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster04\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster04\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster05\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster05\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster05\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster06\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster06\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster06\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster07\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster07\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster07\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster08\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster08\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster08\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-cluster09\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-cluster09\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 29, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-cluster09\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "region:us" ]

[]

[]

[ "## Student Scores Dataset\n\nThis dataset contains clean and original versions of Student Scores Dataset and the transformer used to transform it from original to clean, can be used for inferences.\n\nHere's the plot of the transformer:\n\n<style>#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 {color: black;background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 pre{padding: 0;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable {background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__label-arrow:hover:before {color: black;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator:hover URL-toggleable__label-arrow:before {color: black;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__control:checked~URL-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__control:checked~URL-toggleable__label-arrow:before {content: \"▾\";}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator URL-toggleable__control:checked~URL-toggleable__label {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label URL-toggleable__control:checked~URL-toggleable__label {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator:hover {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label:hover URL-toggleable__label {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 2em;bottom: 0;left: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-item {z-index: 1;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 2em;bottom: 0;left: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item {display: flex;flex-direction: column;position: relative;background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item:only-child::after {width: 0;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;position: relative;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label label {font-family: monospace;font-weight: bold;background-color: white;display: inline-block;line-height: 1.2em;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label-container {position: relative;z-index: 2;text-align: center;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-container {/* jupyter's 'URL' sets '[hidden] { display: none; }' but URL set '[hidden] { display: none !important; }' so we also need the '!important' here to be able to override the default hidden behavior on the sphinx rendered URL. See: URL */display: inline-block !important;position: relative;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-text-repr-fallback {display: none;}</style><div id=\"sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>ColumnTransformer(remainder=&#x27;passthrough&#x27;,transformers=[(&#x27;categorical_missing_value_imputer&#x27;,SimpleImputer(fill_value=&#x27;missing&#x27;,strategy=&#x27;constant&#x27;),[0, 1, 2, 3, 4]),(&#x27;numerical_missing_value_imputer&#x27;,SimpleImputer(strategy=&#x27;median&#x27;), [5, 6, 7]),(&#x27;school_encoder&#x27;, OrdinalEncoder(), [2]),(&#x27;status_encoder&#x27;, OrdinalEncoder(), [4]),(&#x27;gender_encoder&#x27;, OneHotEncoder(), [0])])</pre><b>Please rerun this cell to show the HTML repr or trust the notebook.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"c04042d6-1013-4e6e-97d5-80229d8d887c\" type=\"checkbox\" ><label for=\"c04042d6-1013-4e6e-97d5-80229d8d887c\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">ColumnTransformer</label><div class=\"sk-toggleable__content\"><pre>ColumnTransformer(remainder=&#x27;passthrough&#x27;,transformers=[(&#x27;categorical_missing_value_imputer&#x27;,SimpleImputer(fill_value=&#x27;missing&#x27;,strategy=&#x27;constant&#x27;),[0, 1, 2, 3, 4]),(&#x27;numerical_missing_value_imputer&#x27;,SimpleImputer(strategy=&#x27;median&#x27;), [5, 6, 7]),(&#x27;school_encoder&#x27;, OrdinalEncoder(), [2]),(&#x27;status_encoder&#x27;, OrdinalEncoder(), [4]),(&#x27;gender_encoder&#x27;, OneHotEncoder(), [0])])</pre></div></div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"09be6c7a-7620-4240-ae3e-fea9b9c4ba96\" type=\"checkbox\" ><label for=\"09be6c7a-7620-4240-ae3e-fea9b9c4ba96\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">categorical_missing_value_imputer</label><div class=\"sk-toggleable__content\"><pre>[0, 1, 2, 3, 4]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"26c15d8d-4a1f-4c4d-b0de-5385845dad87\" type=\"checkbox\" ><label for=\"26c15d8d-4a1f-4c4d-b0de-5385845dad87\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">SimpleImputer</label><div class=\"sk-toggleable__content\"><pre>SimpleImputer(fill_value=&#x27;missing&#x27;, strategy=&#x27;constant&#x27;)</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"240be745-c3b3-4b4a-825b-2d1fdb4098c4\" type=\"checkbox\" ><label for=\"240be745-c3b3-4b4a-825b-2d1fdb4098c4\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">numerical_missing_value_imputer</label><div class=\"sk-toggleable__content\"><pre>[5, 6, 7]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"27c7042f-3ced-4afc-ac3a-08b18ef36baa\" type=\"checkbox\" ><label for=\"27c7042f-3ced-4afc-ac3a-08b18ef36baa\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">SimpleImputer</label><div class=\"sk-toggleable__content\"><pre>SimpleImputer(strategy=&#x27;median&#x27;)</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"78993eb3-7988-4fb6-b8e2-c05be3457d30\" type=\"checkbox\" ><label for=\"78993eb3-7988-4fb6-b8e2-c05be3457d30\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">school_encoder</label><div class=\"sk-toggleable__content\"><pre>[2]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"bc1fd86e-4a3b-4448-85d5-15961983cfa2\" type=\"checkbox\" ><label for=\"bc1fd86e-4a3b-4448-85d5-15961983cfa2\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">OrdinalEncoder</label><div class=\"sk-toggleable__content\"><pre>OrdinalEncoder()</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"56bbc2fd-309f-40fc-b160-45fc33cea93b\" type=\"checkbox\" ><label for=\"56bbc2fd-309f-40fc-b160-45fc33cea93b\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">status_encoder</label><div class=\"sk-toggleable__content\"><pre>[4]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"b80005c6-2fe9-4168-971f-8951bfa7f8f3\" type=\"checkbox\" ><label for=\"b80005c6-2fe9-4168-971f-8951bfa7f8f3\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">OrdinalEncoder</label><div class=\"sk-toggleable__content\"><pre>OrdinalEncoder()</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"677cf14a-996a-48af-ba0e-e3d2e83021b8\" type=\"checkbox\" ><label for=\"677cf14a-996a-48af-ba0e-e3d2e83021b8\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">gender_encoder</label><div class=\"sk-toggleable__content\"><pre>[0]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"0cad3051-c4b7-41a8-a372-c439ae4ad98b\" type=\"checkbox\" ><label for=\"0cad3051-c4b7-41a8-a372-c439ae4ad98b\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">OneHotEncoder</label><div class=\"sk-toggleable__content\"><pre>OneHotEncoder()</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"e5707a95-9465-439b-ae0b-34e122add191\" type=\"checkbox\" ><label for=\"e5707a95-9465-439b-ae0b-34e122add191\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">remainder</label><div class=\"sk-toggleable__content\"><pre>[]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"534f7a9b-d224-476c-993a-124b3435a8e3\" type=\"checkbox\" ><label for=\"534f7a9b-d224-476c-993a-124b3435a8e3\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">passthrough</label><div class=\"sk-toggleable__content\"><pre>passthrough</pre></div></div></div></div></div></div></div></div></div></div>" ]

[ "TAGS\n#region-us \n", "## Student Scores Dataset\n\nThis dataset contains clean and original versions of Student Scores Dataset and the transformer used to transform it from original to clean, can be used for inferences.\n\nHere's the plot of the transformer:\n\n<style>#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 {color: black;background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 pre{padding: 0;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable {background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__label-arrow:hover:before {color: black;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator:hover URL-toggleable__label-arrow:before {color: black;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__control:checked~URL-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-toggleable__control:checked~URL-toggleable__label-arrow:before {content: \"▾\";}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator URL-toggleable__control:checked~URL-toggleable__label {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label URL-toggleable__control:checked~URL-toggleable__label {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-estimator:hover {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label:hover URL-toggleable__label {background-color: #d4ebff;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 2em;bottom: 0;left: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-item {z-index: 1;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 2em;bottom: 0;left: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item {display: flex;flex-direction: column;position: relative;background-color: white;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-parallel-item:only-child::after {width: 0;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;position: relative;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label label {font-family: monospace;font-weight: bold;background-color: white;display: inline-block;line-height: 1.2em;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-label-container {position: relative;z-index: 2;text-align: center;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-container {/* jupyter's 'URL' sets '[hidden] { display: none; }' but URL set '[hidden] { display: none !important; }' so we also need the '!important' here to be able to override the default hidden behavior on the sphinx rendered URL. See: URL */display: inline-block !important;position: relative;}#sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949 URL-text-repr-fallback {display: none;}</style><div id=\"sk-46a90950-7a65-4bd5-81b7-b0c3bf7aa949\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>ColumnTransformer(remainder=&#x27;passthrough&#x27;,transformers=[(&#x27;categorical_missing_value_imputer&#x27;,SimpleImputer(fill_value=&#x27;missing&#x27;,strategy=&#x27;constant&#x27;),[0, 1, 2, 3, 4]),(&#x27;numerical_missing_value_imputer&#x27;,SimpleImputer(strategy=&#x27;median&#x27;), [5, 6, 7]),(&#x27;school_encoder&#x27;, OrdinalEncoder(), [2]),(&#x27;status_encoder&#x27;, OrdinalEncoder(), [4]),(&#x27;gender_encoder&#x27;, OneHotEncoder(), [0])])</pre><b>Please rerun this cell to show the HTML repr or trust the notebook.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"c04042d6-1013-4e6e-97d5-80229d8d887c\" type=\"checkbox\" ><label for=\"c04042d6-1013-4e6e-97d5-80229d8d887c\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">ColumnTransformer</label><div class=\"sk-toggleable__content\"><pre>ColumnTransformer(remainder=&#x27;passthrough&#x27;,transformers=[(&#x27;categorical_missing_value_imputer&#x27;,SimpleImputer(fill_value=&#x27;missing&#x27;,strategy=&#x27;constant&#x27;),[0, 1, 2, 3, 4]),(&#x27;numerical_missing_value_imputer&#x27;,SimpleImputer(strategy=&#x27;median&#x27;), [5, 6, 7]),(&#x27;school_encoder&#x27;, OrdinalEncoder(), [2]),(&#x27;status_encoder&#x27;, OrdinalEncoder(), [4]),(&#x27;gender_encoder&#x27;, OneHotEncoder(), [0])])</pre></div></div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"09be6c7a-7620-4240-ae3e-fea9b9c4ba96\" type=\"checkbox\" ><label for=\"09be6c7a-7620-4240-ae3e-fea9b9c4ba96\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">categorical_missing_value_imputer</label><div class=\"sk-toggleable__content\"><pre>[0, 1, 2, 3, 4]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"26c15d8d-4a1f-4c4d-b0de-5385845dad87\" type=\"checkbox\" ><label for=\"26c15d8d-4a1f-4c4d-b0de-5385845dad87\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">SimpleImputer</label><div class=\"sk-toggleable__content\"><pre>SimpleImputer(fill_value=&#x27;missing&#x27;, strategy=&#x27;constant&#x27;)</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"240be745-c3b3-4b4a-825b-2d1fdb4098c4\" type=\"checkbox\" ><label for=\"240be745-c3b3-4b4a-825b-2d1fdb4098c4\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">numerical_missing_value_imputer</label><div class=\"sk-toggleable__content\"><pre>[5, 6, 7]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"27c7042f-3ced-4afc-ac3a-08b18ef36baa\" type=\"checkbox\" ><label for=\"27c7042f-3ced-4afc-ac3a-08b18ef36baa\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">SimpleImputer</label><div class=\"sk-toggleable__content\"><pre>SimpleImputer(strategy=&#x27;median&#x27;)</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"78993eb3-7988-4fb6-b8e2-c05be3457d30\" type=\"checkbox\" ><label for=\"78993eb3-7988-4fb6-b8e2-c05be3457d30\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">school_encoder</label><div class=\"sk-toggleable__content\"><pre>[2]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"bc1fd86e-4a3b-4448-85d5-15961983cfa2\" type=\"checkbox\" ><label for=\"bc1fd86e-4a3b-4448-85d5-15961983cfa2\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">OrdinalEncoder</label><div class=\"sk-toggleable__content\"><pre>OrdinalEncoder()</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"56bbc2fd-309f-40fc-b160-45fc33cea93b\" type=\"checkbox\" ><label for=\"56bbc2fd-309f-40fc-b160-45fc33cea93b\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">status_encoder</label><div class=\"sk-toggleable__content\"><pre>[4]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"b80005c6-2fe9-4168-971f-8951bfa7f8f3\" type=\"checkbox\" ><label for=\"b80005c6-2fe9-4168-971f-8951bfa7f8f3\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">OrdinalEncoder</label><div class=\"sk-toggleable__content\"><pre>OrdinalEncoder()</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"677cf14a-996a-48af-ba0e-e3d2e83021b8\" type=\"checkbox\" ><label for=\"677cf14a-996a-48af-ba0e-e3d2e83021b8\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">gender_encoder</label><div class=\"sk-toggleable__content\"><pre>[0]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"0cad3051-c4b7-41a8-a372-c439ae4ad98b\" type=\"checkbox\" ><label for=\"0cad3051-c4b7-41a8-a372-c439ae4ad98b\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">OneHotEncoder</label><div class=\"sk-toggleable__content\"><pre>OneHotEncoder()</pre></div></div></div></div></div></div><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"e5707a95-9465-439b-ae0b-34e122add191\" type=\"checkbox\" ><label for=\"e5707a95-9465-439b-ae0b-34e122add191\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">remainder</label><div class=\"sk-toggleable__content\"><pre>[]</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"534f7a9b-d224-476c-993a-124b3435a8e3\" type=\"checkbox\" ><label for=\"534f7a9b-d224-476c-993a-124b3435a8e3\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">passthrough</label><div class=\"sk-toggleable__content\"><pre>passthrough</pre></div></div></div></div></div></div></div></div></div></div>" ]

[ 6, 4951 ]

[ "passage: TAGS\n#region-us \n" ]

[ "task_categories:sentence-similarity", "task_ids:semantic-similarity-classification", "language:en", "license:mit", "region:us" ]

[]

[ "en" ]

[ "# Dataset Card for \"WikiAnswers\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Anthony Fader, Luke Zettlemoyer, Oren Etzioni", "### Dataset Summary\nThe WikiAnswers corpus contains clusters of questions tagged by WikiAnswers users as paraphrases. \nEach cluster optionally contains an answer provided by WikiAnswers users. There are 30,370,994 clusters containing an average of 25 questions per cluster. 3,386,256 (11%) of the clusters have an answer.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains 25 equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format for 'N' examples:\n\nReview an example 'i' with:", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ "TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n", "# Dataset Card for \"WikiAnswers\"", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions", "## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Anthony Fader, Luke Zettlemoyer, Oren Etzioni", "### Dataset Summary\nThe WikiAnswers corpus contains clusters of questions tagged by WikiAnswers users as paraphrases. \nEach cluster optionally contains an answer provided by WikiAnswers users. There are 30,370,994 clusters containing an average of 25 questions per cluster. 3,386,256 (11%) of the clusters have an answer.", "### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.", "### Languages\n- English.", "## Dataset Structure\nEach example in the dataset contains 25 equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.", "### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format for 'N' examples:\n\nReview an example 'i' with:", "### Data Instances", "### Data Fields", "### Data Splits", "## Dataset Creation", "### Curation Rationale", "### Source Data", "#### Initial Data Collection and Normalization", "#### Who are the source language producers?", "### Annotations", "#### Annotation process", "#### Who are the annotators?", "### Personal and Sensitive Information", "## Considerations for Using the Data", "### Social Impact of Dataset", "### Discussion of Biases", "### Other Known Limitations", "## Additional Information", "### Dataset Curators", "### Licensing Information", "### Contributions" ]

[ 43, 11, 120, 35, 81, 24, 7, 74, 65, 6, 5, 5, 5, 7, 4, 10, 10, 5, 5, 9, 8, 8, 7, 8, 7, 5, 6, 6, 5 ]

[ "passage: TAGS\n#task_categories-sentence-similarity #task_ids-semantic-similarity-classification #language-English #license-mit #region-us \n# Dataset Card for \"WikiAnswers\"## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks and Leaderboards\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information\n - Contributions## Dataset Description\n\n- Homepage: URL\n- Repository: \n- Paper: \n- Point of Contact: Anthony Fader, Luke Zettlemoyer, Oren Etzioni### Dataset Summary\nThe WikiAnswers corpus contains clusters of questions tagged by WikiAnswers users as paraphrases. \nEach cluster optionally contains an answer provided by WikiAnswers users. There are 30,370,994 clusters containing an average of 25 questions per cluster. 3,386,256 (11%) of the clusters have an answer.### Supported Tasks\n- Sentence Transformers training; useful for semantic search and sentence similarity.### Languages\n- English.## Dataset Structure\nEach example in the dataset contains 25 equivalent sentences and is formatted as a dictionary with the key \"set\" and a list with the sentences as \"value\".\n\nThis dataset is useful for training Sentence Transformers models. Refer to the following post on how to train models using similar sentences.### Usage Example\nInstall the Datasets library with 'pip install datasets' and load the dataset from the Hub with:\n\nThe dataset is loaded as a 'DatasetDict' and has the format for 'N' examples:\n\nReview an example 'i' with:### Data Instances### Data Fields### Data Splits## Dataset Creation### Curation Rationale### Source Data#### Initial Data Collection and Normalization" ]

[ "annotations_creators:expert-generated", "language_creators:found", "multilinguality:monolingual", "size_categories:n<1K", "source_datasets:original", "language:ch", "license:afl-3.0", "region:us" ]

[]

[ "ch" ]

[]

[ "TAGS\n#annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-n<1K #source_datasets-original #language-Chamorro #license-afl-3.0 #region-us \n" ]

[ 67 ]

[ "passage: TAGS\n#annotations_creators-expert-generated #language_creators-found #multilinguality-monolingual #size_categories-n<1K #source_datasets-original #language-Chamorro #license-afl-3.0 #region-us \n" ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-rated-low\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-rated-low\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 30, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-rated-low\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-rated-medium\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-rated-medium\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 31, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-rated-medium\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]

[ "task_categories:multiple-choice", "task_categories:question-answering", "task_categories:zero-shot-classification", "task_categories:text2text-generation", "task_categories:table-question-answering", "task_categories:text-generation", "task_categories:text-classification", "task_categories:tabular-classification", "task_ids:multiple-choice-qa", "task_ids:extractive-qa", "task_ids:open-domain-qa", "task_ids:closed-domain-qa", "task_ids:closed-book-qa", "task_ids:open-book-qa", "task_ids:language-modeling", "task_ids:multi-class-classification", "task_ids:natural-language-inference", "task_ids:topic-classification", "task_ids:multi-label-classification", "task_ids:tabular-multi-class-classification", "task_ids:tabular-multi-label-classification", "annotations_creators:no-annotation", "language_creators:found", "multilinguality:monolingual", "size_categories:100K<n<1M", "language:en", "license:apache-2.0", "arxiv:2208.01009", "region:us" ]

[ "2208.01009" ]

[ "en" ]

[ "# Dataset Card for \"UnpredicTable-rated-high\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ "TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n", "# Dataset Card for \"UnpredicTable-rated-high\" - Dataset of Few-shot Tasks from Tables", "## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise", "### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.", "### Languages\n\nEnglish", "## Dataset Structure", "### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.", "### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.", "## Dataset Creation", "### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.", "### Source Data", "#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.", "#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.", "### Annotations", "#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.", "#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant.", "### Personal and Sensitive Information\n\nThe data was extracted from WDC Web Table Corpora, which in turn extracted tables from the Common Crawl. We did not filter the data in any way. Thus any user identities or otherwise sensitive information (e.g., data that reveals racial or ethnic origins, sexual orientations, religious beliefs, political opinions or union memberships, or locations; financial or health data; biometric or genetic data; forms of government identification, such as social security numbers; criminal history, etc.) might be contained in our dataset.", "## Considerations for Using the Data", "### Social Impact of Dataset\n\nThis dataset is intended for use as a research resource to investigate the relationship between training data and few-shot learning. As such, it contains high- and low-quality data, as well as diverse content that may be untruthful or inappropriate. Without careful investigation, it should not be used for training models that will be deployed for use in decision-critical or user-facing situations.", "### Discussion of Biases\n\nSince our dataset contains tables that are scraped from the web, it will also contain many toxic, racist, sexist, and otherwise harmful biases and texts. We have not run any analysis on the biases prevalent in our datasets. Neither have we explicitly filtered the content. This implies that a model trained on our dataset may potentially reflect harmful biases and toxic text that exist in our dataset.", "### Other Known Limitations\n\nNo additional known limitations.", "## Additional Information", "### Dataset Curators\nJun Shern Chan, Michael Pieler, Jonathan Jao, Jérémy Scheurer, Ethan Perez", "### Licensing Information\nApache 2.0" ]

[ 324, 30, 112, 47, 866, 169, 5, 6, 205, 122, 23, 5, 141, 4, 109, 23, 5, 72, 20, 130, 8, 97, 112, 13, 5, 30, 9 ]

[ "passage: TAGS\n#task_categories-multiple-choice #task_categories-question-answering #task_categories-zero-shot-classification #task_categories-text2text-generation #task_categories-table-question-answering #task_categories-text-generation #task_categories-text-classification #task_categories-tabular-classification #task_ids-multiple-choice-qa #task_ids-extractive-qa #task_ids-open-domain-qa #task_ids-closed-domain-qa #task_ids-closed-book-qa #task_ids-open-book-qa #task_ids-language-modeling #task_ids-multi-class-classification #task_ids-natural-language-inference #task_ids-topic-classification #task_ids-multi-label-classification #task_ids-tabular-multi-class-classification #task_ids-tabular-multi-label-classification #annotations_creators-no-annotation #language_creators-found #multilinguality-monolingual #size_categories-100K<n<1M #language-English #license-apache-2.0 #arxiv-2208.01009 #region-us \n# Dataset Card for \"UnpredicTable-rated-high\" - Dataset of Few-shot Tasks from Tables## Table of Contents\n- Dataset Description\n - Dataset Summary\n - Supported Tasks\n - Languages\n- Dataset Structure\n - Data Instances\n - Data Fields\n - Data Splits\n- Dataset Creation\n - Curation Rationale\n - Source Data\n - Annotations\n - Personal and Sensitive Information\n- Considerations for Using the Data\n - Social Impact of Dataset\n - Discussion of Biases\n - Other Known Limitations\n- Additional Information\n - Dataset Curators\n - Licensing Information\n - Citation Information", "passage: ## Dataset Description\n\n- Homepage: URL\n- Repository: URL\n- Paper: Few-shot Adaptation Works with UnpredicTable Data\n- Point of Contact: junshern@URL, perez@URL", "passage: ### Dataset Summary\n\nThe UnpredicTable dataset consists of web tables formatted as few-shot tasks for fine-tuning language models to improve their few-shot performance.\n\nThere are several dataset versions available:\n\n* UnpredicTable-full: Starting from the initial WTC corpus of 50M tables, we apply our tables-to-tasks procedure to produce our resulting dataset, UnpredicTable-full, which comprises 413,299 tasks from 23,744 unique websites.\n\n* UnpredicTable-unique: This is the same as UnpredicTable-full but filtered to have a maximum of one task per website. UnpredicTable-unique contains exactly 23,744 tasks from 23,744 websites.\n\n* UnpredicTable-5k: This dataset contains 5k random tables from the full dataset.\n\n* UnpredicTable data subsets based on a manual human quality rating (please see our publication for details of the ratings):\n * UnpredicTable-rated-low\n * UnpredicTable-rated-medium\n * UnpredicTable-rated-high\n\n* UnpredicTable data subsets based on the website of origin:\n * UnpredicTable-baseball-fantasysports-yahoo-com \n * UnpredicTable-bulbapedia-bulbagarden-net\n * UnpredicTable-cappex-com \n * UnpredicTable-cram-com\n * UnpredicTable-dividend-com \n * UnpredicTable-dummies-com\n * UnpredicTable-en-wikipedia-org\n * UnpredicTable-ensembl-org\n * UnpredicTable-gamefaqs-com\n * UnpredicTable-mgoblog-com\n * UnpredicTable-mmo-champion-com\n * UnpredicTable-msdn-microsoft-com\n * UnpredicTable-phonearena-com\n * UnpredicTable-sittercity-com\n * UnpredicTable-sporcle-com\n * UnpredicTable-studystack-com\n * UnpredicTable-support-google-com\n * UnpredicTable-w3-org\n * UnpredicTable-wiki-openmoko-org\n * UnpredicTable-wkdu-org\n\n\n* UnpredicTable data subsets based on clustering (for the clustering details please see our publication):\n * UnpredicTable-cluster00\n * UnpredicTable-cluster01\n * UnpredicTable-cluster02\n * UnpredicTable-cluster03\n * UnpredicTable-cluster04\n * UnpredicTable-cluster05\n * UnpredicTable-cluster06\n * UnpredicTable-cluster07\n * UnpredicTable-cluster08\n * UnpredicTable-cluster09\n * UnpredicTable-cluster10\n * UnpredicTable-cluster11\n * UnpredicTable-cluster12\n * UnpredicTable-cluster13\n * UnpredicTable-cluster14\n * UnpredicTable-cluster15\n * UnpredicTable-cluster16\n * UnpredicTable-cluster17\n * UnpredicTable-cluster18\n * UnpredicTable-cluster19\n * UnpredicTable-cluster20\n * UnpredicTable-cluster21\n * UnpredicTable-cluster22\n * UnpredicTable-cluster23\n * UnpredicTable-cluster24\n * UnpredicTable-cluster25\n * UnpredicTable-cluster26\n * UnpredicTable-cluster27\n * UnpredicTable-cluster28\n * UnpredicTable-cluster29\n * UnpredicTable-cluster-noise### Supported Tasks and Leaderboards\n\nSince the tables come from the web, the distribution of tasks and topics is very broad. The shape of our dataset is very wide, i.e., we have 1000's of tasks, while each task has only a few examples, compared to most current NLP datasets which are very deep, i.e., 10s of tasks with many examples. This implies that our dataset covers a broad range of potential tasks, e.g., multiple-choice, question-answering, table-question-answering, text-classification, etc.\n\nThe intended use of this dataset is to improve few-shot performance by fine-tuning/pre-training on our dataset.### Languages\n\nEnglish## Dataset Structure### Data Instances\n\nEach task is represented as a jsonline file and consists of several few-shot examples. Each example is a dictionary containing a field 'task', which identifies the task, followed by an 'input', 'options', and 'output' field. The 'input' field contains several column elements of the same row in the table, while the 'output' field is a target which represents an individual column of the same row. Each task contains several such examples which can be concatenated as a few-shot task. In the case of multiple choice classification, the 'options' field contains the possible classes that a model needs to choose from.\n\nThere are also additional meta-data fields such as 'pageTitle', 'title', 'outputColName', 'url', 'wdcFile'.### Data Fields\n\n'task': task identifier\n\n'input': column elements of a specific row in the table. \n\n'options': for multiple choice classification, it provides the options to choose from.\n\n'output': target column element of the same row as input.\n\n'pageTitle': the title of the page containing the table. \n\n'outputColName': output column name\n\n'url': url to the website containing the table\n\n'wdcFile': WDC Web Table Corpus file", "passage: ### Data Splits\n\nThe UnpredicTable datasets do not come with additional data splits.## Dataset Creation### Curation Rationale\n\nFew-shot training on multi-task datasets has been demonstrated to improve language models' few-shot learning (FSL) performance on new tasks, but it is unclear which training tasks lead to effective downstream task adaptation. Few-shot learning datasets are typically produced with expensive human curation, limiting the scale and diversity of the training tasks available to study. As an alternative source of few-shot data, we automatically extract 413,299 tasks from diverse internet tables. We provide this as a research resource to investigate the relationship between training data and few-shot learning.### Source Data#### Initial Data Collection and Normalization\n\nWe use internet tables from the English-language Relational Subset of the WDC Web Table Corpus 2015 (WTC). The WTC dataset tables were extracted from the July 2015 Common Crawl web corpus (URL The dataset contains 50,820,165 tables from 323,160 web domains. We then convert the tables into few-shot learning tasks. Please see our publication for more details on the data collection and conversion pipeline.#### Who are the source language producers?\n\nThe dataset is extracted from WDC Web Table Corpora.### Annotations#### Annotation process\n\nManual annotation was only carried out for the UnpredicTable-rated-low,\nUnpredicTable-rated-medium, and UnpredicTable-rated-high data subsets to rate task quality. Detailed instructions of the annotation instructions can be found in our publication.#### Who are the annotators?\n\nAnnotations were carried out by a lab assistant." ]