Spaces:
Runtime error
Runtime error
File size: 4,199 Bytes
97aff81 3606254 6425dcc 3606254 d7cb910 6425dcc d7cb910 3606254 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 |
import gradio as gr
from frame_semantic_transformer import FrameSemanticTransformer
from transformers import T5ForConditionalGeneration, T5TokenizerFast
from frame_semantic_transformer.data.frame_types import Frame
from frame_semantic_transformer.data.loaders.loader import InferenceLoader
import pandas as pd
import ast
import re
import os
from nltk.stem import SnowballStemmer
# Huggingface model path
huggingface_model_path = "nelsonjq/frame-semantic-transformer-french-small"
# Load the model and tokenizer from Huggingface
model = T5ForConditionalGeneration.from_pretrained(huggingface_model_path)
tokenizer = T5TokenizerFast.from_pretrained(huggingface_model_path)
# Load the DataFrame
import subprocess
os.makedirs("Asfalda", exist_ok=True)
subprocess.run(['wget', '--output-document=Asfalda/frame_lus_df.tsv', 'https://seafile.unistra.fr/f/0155ced00b8d441eb131/?dl=1']
)
frame_lus_df = pd.read_csv("Asfalda/frame_lus_df.tsv", delimiter='\t')
# Filter out 'Other_sense' and normalize frame names
frame_lus_df = frame_lus_df[frame_lus_df['Name'] != 'Other_sense']
frame_lus_df['Name'] = frame_lus_df['Name'].apply(lambda x: x.split(".")[0])
frame_lus_df['Name'] = frame_lus_df['Name'].str.replace(r'^[Ff][Rr][Vv]_', '', regex=True)
# Add missing frame names
new_row_Suasion = {
'Name': 'Suasion',
'Core_Elms': "['Content', 'Cognizer', 'Persuader', 'Target', 'Text', 'Action', 'Addressee']",
'Non_Core_Elms': "['Speaker', 'Topic']",
'Lus': "['convaincre.v', 'convertir.v', 'persuader.v', 'convaincant.a', 'persuasion.n', 'dissuader.v', 'apprendre.v', 'dissuasion.n', 'décider.v']",
'Lus_simple': "['convaincre', 'convertir', 'persuader', 'convaincant', 'persuasion', 'dissuader', 'apprendre', 'dissuasion', 'décider']"
}
new_row_Arriving = {
'Name': 'Arriving',
'Core_Elms': "['Theme', 'Goal']",
'Non_Core_Elms': "['Target', 'Means']",
'Lus': "['gagner.v']",
'Lus_simple': "['gagner']"
}
new_row_Suasion_df = pd.DataFrame([new_row_Suasion])
new_row_Arriving_df = pd.DataFrame([new_row_Arriving])
frame_lus_df = pd.concat([frame_lus_df, new_row_Suasion_df, new_row_Arriving_df], ignore_index=True)
# Define FrenchInferenceLoader
french_stemmer = SnowballStemmer("french")
def extract_frame(df_row_frame) -> Frame:
name = df_row_frame['Name']
core_elms = ast.literal_eval(df_row_frame['Core_Elms'])
non_core_elms = ast.literal_eval(df_row_frame['Non_Core_Elms'])
lus = ast.literal_eval(df_row_frame['Lus'])
return Frame(name=name, core_elements=core_elms, non_core_elements=non_core_elms, lexical_units=lus)
class FrenchInferenceLoader(InferenceLoader):
def __init__(self, french_framenet_df_file):
self.frames = []
for index, row in french_framenet_df_file.iterrows():
frame = extract_frame(row)
self.frames.append(frame)
def load_frames(self):
return self.frames
def normalize_lexical_unit_text(self, lu: str) -> str:
normalized_lu = lu.lower()
if '.' in normalized_lu:
normalized_lu = normalized_lu.split('.')[0]
normalized_lu = re.sub(r"[^a-z0-9 ]", "", normalized_lu)
return french_stemmer.stem(normalized_lu)
# Initialize the FrenchInferenceLoader and FrameSemanticTransformer
inference_loader = FrenchInferenceLoader(frame_lus_df)
transformer = FrameSemanticTransformer(huggingface_model_path, inference_loader=inference_loader)
# Function to process the input sentence and display frame detection results
def detect_frames_in_text(input_text):
result = transformer.detect_frames(input_text)
output = f"Results found in the sentence:\n\n{result.sentence}\n"
for frame in result.frames:
output += f"\nFRAME: {frame.name}\n\nFrame Elements:\n"
for element in frame.frame_elements:
output += f"\t\t{element.name}: {element.text}\n"
return output
# Gradio interface
iface = gr.Interface(
fn=detect_frames_in_text,
inputs="text",
outputs="text",
title="French Frame Detection App",
description="Enter a French sentence to detect frames and frame elements using the FrameSemanticTransformer model."
)
# Launch the app
iface.launch()
|