Add Darija transcription and topic extraction app8

app.py

@@ -3,20 +3,24 @@ import torch
 from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC, pipeline
 from transformers import BertTokenizer, BertForSequenceClassification
 import librosa
+import os
+
+# Set up proxy for internal testing
+os.environ["HTTP_PROXY"] = "http://meditelproxy.meditel.int:80"
+os.environ["HTTPS_PROXY"] = "http://meditelproxy.meditel.int:80"
 
 # Load models
 # Transcription model for Moroccan Darija
 processor = Wav2Vec2Processor.from_pretrained("boumehdi/wav2vec2-large-xlsr-moroccan-darija")
 transcription_model = Wav2Vec2ForCTC.from_pretrained("boumehdi/wav2vec2-large-xlsr-moroccan-darija")
 
-# Summarization model
-summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
+# Summarization model (for French summaries)
+summarizer = pipeline("summarization", model="facebook/mbart-large-50-many-to-many-mmt")
 
 # Topic Classification Model (BERT for example)
 topic_model = BertForSequenceClassification.from_pretrained("bert-base-uncased")  # Example model
 topic_tokenizer = BertTokenizer.from_pretrained("bert-base-uncased")
 
-
 # Function to resample audio to 16kHz if necessary
 def resample_audio(audio_path, target_sr=16000):
     audio_input, original_sr = librosa.load(audio_path, sr=None)  # Load audio with original sampling rate
@@ -57,13 +61,19 @@ def classify_topic(transcription):
     else:
         return "Other"
 
-# Function to transcribe, summarize, and classify topic
+# Function to transcribe, summarize in French, and classify topic
 def transcribe_and_summarize(audio_file):
     # Transcription
     transcription = transcribe_audio(audio_file)
 
-    # Summarization
-    summary = summarizer(transcription, max_length=50, min_length=10, do_sample=False)[0]["summary_text"]
+    # Summarization in French
+    summary = summarizer(
+        transcription,
+        max_length=50,
+        min_length=10,
+        do_sample=False,
+        tgt_lang="fr_XX"  # Target language set to French
+    )[0]["summary_text"]
 
     # Topic classification
     topic = classify_topic(transcription)
@@ -74,7 +84,7 @@ def transcribe_and_summarize(audio_file):
 inputs = gr.Audio(type="filepath", label="Upload your audio file")
 outputs = [
     gr.Textbox(label="Transcription"),
-    gr.Textbox(label="Summary"),
+    gr.Textbox(label="Résumé (en Français)"),
     gr.Textbox(label="Topic")
 ]
 
@@ -82,8 +92,8 @@ app = gr.Interface(
     fn=transcribe_and_summarize,
     inputs=inputs,
     outputs=outputs,
-    title="Moroccan Darija Audio Transcription, Summarization, and Topic Classification",
-    description="Upload an audio file in Moroccan Darija to get its transcription, a summarized version of the content, and the detected topic."
+    title="Moroccan Darija Audio Transcription, Résumé, and Topic Classification",
+    description="Upload an audio file in Moroccan Darija to get its transcription, a summarized version in French, and the detected topic."
 )
 
 # Launch the app