emotion_detection.py

from transformers import AutoTokenizer, AutoModelForSequenceClassification
from transformers_interpret import SequenceClassificationExplainer
import torch
import pandas as pd

class EmotionDetection:
    """
    Emotion Detection on text data.
    Attributes:
        tokenizer: An instance of Hugging Face Tokenizer
        model: An instance of Hugging Face Model
        explainer: An instance of SequenceClassificationExplainer from Transformers interpret
    """

    def __init__(self):
        hub_location = 'cardiffnlp/twitter-roberta-base-emotion'
        self.tokenizer = AutoTokenizer.from_pretrained(hub_location)
        self.model = AutoModelForSequenceClassification.from_pretrained(hub_location)
        self.explainer = SequenceClassificationExplainer(self.model, self.tokenizer)

    def justify(self, text):
        """
        Get html annotation for displaying emotion justification over text.
        Parameters:
            text (str): The user input string to emotion justification
        Returns:
            html (html): html object for plotting emotion prediction justification
        """
        word_attributions = self.explainer(text)
        html = self.explainer.visualize("example.html")
        return html

    def classify(self, text):
        """
        Recognize Emotion in text.
        Parameters:
            text (str): The user input string to perform emotion classification on
        Returns:
            predictions (str): The predicted probabilities for emotion classes
        """

        tokens = self.tokenizer.encode_plus(text, add_special_tokens=False, return_tensors='pt')
        outputs = self.model(**tokens)
        probs = torch.nn.functional.softmax(outputs[0], dim=-1)
        probs = probs.mean(dim=0).detach().numpy()

        # Original labels from model
        original_labels = list(self.model.config.id2label.values())

        # Only keep the 4 specific emotions and map custom names
        desired_labels = ['joy', 'anger', 'sadness', 'optimism']
        custom_labels = {
            'joy': 'Happiness 😊',
            'anger': 'Anger 😡',
            'sadness': 'Sadness 😢',
            'optimism': 'Hopeful ✨'
        }

        filtered_probs = []
        filtered_labels = []

        for label, prob in zip(original_labels, probs):
            if label in desired_labels:
                filtered_probs.append(prob)
                filtered_labels.append(custom_labels[label])

        preds = pd.Series(filtered_probs, index=filtered_labels, name='Predicted Probability')
        return preds

    def run(self, text):
        """
        Classify and Justify Emotion in text.
        Parameters:
            text (str): The user input string to perform emotion classification on
        Returns:
            predictions (str): The predicted probabilities for emotion classes
            html (html): html object for plotting emotion prediction justification
        """
        preds = self.classify(text)
        html = self.justify(text)
        return preds, html