Update whisper.py

whisper.py

@@ -29,19 +29,8 @@ pipe = pipeline(
     token=os.getenv("HF_TOKEN")
     ) 
 
-def post_process_transcription(example_transcription, max_repeats=1):
-    segments = re.findall(r'.+?[.,?]', example_transcription)
-
-    seen = set()
-    unique_segments = []
-    for segment in segments:
-        if segment not in seen:
-            unique_segments.append(segment)
-            seen.add(segment)
-
-    final_string = ''.join(unique_segments)
-
-    tokens = re.findall(r'\b\w+\b[.,!?]?', final_string)
+def post_process_transcription(transcription, max_repeats=2):
+    tokens = re.findall(r'\b\w+\'?\w*\b[.,!?]?', transcription)
 
     cleaned_tokens = []
     repetition_count = 0
@@ -65,6 +54,7 @@ def post_process_transcription(example_transcription, max_repeats=1):
 
     return cleaned_transcription
 
+
 def convert_forced_to_tokens(forced_decoder_ids):
     forced_decoder_tokens = []
     for i, (idx, token) in enumerate(forced_decoder_ids):
@@ -236,6 +226,4 @@ def generate(audio_path, use_v5):
         
     clean_output = post_process_transcription(output, max_repeats=1)
 
-    return clean_output 
-
-        
+    return clean_output