Update app.py

app.py

@@ -1,6 +1,7 @@
 import gradio as gr
 import torch
 from transformers import pipeline
+from jiwer import wer
 
 # Load models
 whisper_pipeline_1 = pipeline("automatic-speech-recognition", model="maliahson/whisper-agri")
@@ -8,23 +9,42 @@ device = 0 if torch.cuda.is_available() else "cpu"
 whisper_pipeline_2 = pipeline("automatic-speech-recognition", model="openai/whisper-large-v3-turbo", device=device)
 
 def transcribe_and_compare(audio_path, original_transcription=None):
+    """
+    Transcribes an audio file using two Whisper models and compares results.
+    
+    Args:
+        audio_path (str): Path to the audio file.
+        original_transcription (str, optional): Ground truth transcription.
+
+    Returns:
+        dict: Results including transcriptions and WER calculations.
+    """
     transcription_1 = whisper_pipeline_1(audio_path)["text"]
     transcription_2 = whisper_pipeline_2(audio_path)["text"]
-    comparison_result = {"Model 1 Output": transcription_1, "Model 2 Output": transcription_2}
+    comparison_result = {
+        "Model 1 Output (maliahson/whisper-agri)": transcription_1,
+        "Model 2 Output (openai/whisper-large-v3-turbo)": transcription_2,
+    }
 
     if original_transcription:
-        from jiwer import wer
+        # Calculate Word Error Rate
         wer_1 = wer(original_transcription, transcription_1)
         wer_2 = wer(original_transcription, transcription_2)
         comparison_result["WER Model 1"] = wer_1
         comparison_result["WER Model 2"] = wer_2
+    else:
+        # Compare outputs of both models
+        comparison_result["Difference Between Models"] = {
+            "Model 1 Unique Words": set(transcription_1.split()) - set(transcription_2.split()),
+            "Model 2 Unique Words": set(transcription_2.split()) - set(transcription_1.split()),
+        }
 
     return comparison_result
 
 # Gradio Interface
 with gr.Blocks() as demo:
     gr.Markdown("## Audio Transcription and Comparison")
-    audio_input = gr.Audio(type="filepath", label="Upload or Record Audio (supports upload and recording)")
+    audio_input = gr.Audio(type="filepath", label="Upload or Record Audio")
     original_transcription = gr.Textbox(lines=2, label="Original Transcription (Optional)")
     output = gr.JSON(label="Comparison Results")
     submit_btn = gr.Button("Transcribe and Compare")
@@ -35,4 +55,4 @@ with gr.Blocks() as demo:
         outputs=output
     )
 
-demo.launch(debug=True)
+demo.launch(debug=True)