Update app.py

app.py

@@ -34,38 +34,38 @@ def parse_transcription(logits):
     return transcription
 
 
-# def parse(wav_file):
-#     input_values = read_file_and_process(wav_file)
-#     with torch.no_grad():
-#         logits = model(**input_values).logits
-#     user_question = parse_transcription(logits)
-#     return user_question
+def parse(wav_file):
+    input_values = read_file_and_process(wav_file)
+    with torch.no_grad():
+        logits = model(**input_values).logits
+    user_question = parse_transcription(logits)
+    return user_question
 
 
 # Function to retrieve an answer based on a question (using fuzzy matching)
-def get_answer(wav_file=None, text=None):
+# def get_answer(wav_file=None, text=None):
     
-    if type(wav_file) != 'str' or type(text != 'str'):
-        input_values = read_file_and_process(wav_file)
-        with torch.no_grad():
-            logits = model(**input_values).logits
-        user_question = parse_transcription(logits)
-    else:
-        user_question = wav_file
+#     if type(wav_file) != 'str' or type(text != 'str'):
+#         input_values = read_file_and_process(wav_file)
+#         with torch.no_grad():
+#             logits = model(**input_values).logits
+#         user_question = parse_transcription(logits)
+#     else:
+#         user_question = wav_file
         
-    highest_score = 0
-    best_answer = None
+#     highest_score = 0
+#     best_answer = None
 
-    for item in DATASET:
-        similarity_score = fuzz.token_set_ratio(user_question, item["question"])
-        if similarity_score > highest_score:
-            highest_score = similarity_score
-            best_answer = item["answer"]
+#     for item in DATASET:
+#         similarity_score = fuzz.token_set_ratio(user_question, item["question"])
+#         if similarity_score > highest_score:
+#             highest_score = similarity_score
+#             best_answer = item["answer"]
 
-    if highest_score >= 80:  # Adjust the similarity threshold as needed
-        return best_answer
-    else:
-        return "I don't have an answer to that question."
+#     if highest_score >= 80:  # Adjust the similarity threshold as needed
+#         return best_answer
+#     else:
+#         return "I don't have an answer to that question."
 
 
 model_id = "jonatasgrosman/wav2vec2-large-xlsr-53-persian"
@@ -79,10 +79,10 @@ input_ = [
                   show_download_button=True,
                   show_edit_button=True,
                  ), 
-        gr.Textbox(label="سوال خود را بنویسید.",
-                   lines=3,
-                   text_align="right",
-                   show_label=True,)
+        # gr.Textbox(label="سوال خود را بنویسید.",
+        #            lines=3,
+        #            text_align="right",
+        #            show_label=True,)
          ]
 
 txtbox = gr.Textbox(
@@ -97,7 +97,7 @@ title = "Speech-to-Text (persian)"
 description = "، توجه داشته باشید که هرچه گفتار شما شمرده تر باشد خروجی با کیفیت تری دارید.روی دکمه ضبط صدا کلیک کنید و سپس دسترسی مرورگر خود را به میکروفون دستگاه بدهید، سپس شروع به صحبت کنید و برای اتمام ضبط دوباره روی دکمه کلیک کنید"
 article = "<p style='text-align: center'><a href='https://github.com/nimaprgrmr'>Large-Scale Self- and Semi-Supervised Learning for Speech Translation</a></p>"
 
-demo = gr.Interface(fn=get_answer, inputs = input_,  outputs=txtbox, title=title, description=description, article = article,
+demo = gr.Interface(fn=parse, inputs = input_,  outputs=txtbox, title=title, description=description, article = article,
              streaming=True, interactive=True,
              analytics_enabled=False, show_tips=False, enable_queue=True)
 demo.launch(share=True)