Update app.py

app.py

@@ -32,7 +32,7 @@ from transformers import T5ForConditionalGeneration, T5Tokenizer
 __FILES = set()
 wispher_models = list(whisper._MODELS.keys())
 
-def correct_grammar(input_text,num_return_sequences=num_return_sequences):
+def correct_grammar(input_text,num_return_sequences="1"):
     torch_device = 'cuda' if torch.cuda.is_available() else 'cpu'
     tokenizer = T5Tokenizer.from_pretrained('deep-learning-analytics/GrammarCorrector')
     model = T5ForConditionalGeneration.from_pretrained('deep-learning-analytics/GrammarCorrector').to(torch_device)
@@ -42,7 +42,16 @@ def correct_grammar(input_text,num_return_sequences=num_return_sequences):
     for generated_sequence_idx, generated_sequence in enumerate(results):
         text = tokenizer.decode(generated_sequence, clean_up_tokenization_spaces=True, skip_special_tokens=True)
         generated_sequences.append(text)
-    return "".join(generated_sequences)
+    generated_text = "".join(generated_sequences)
+    _generated_text = ""
+    for idx, _sentence in enumerate(generated_text.split('.'), 0):
+        if not idx:
+            _generated_text+=_sentence+'.'
+        elif _sentence[:1]!=' ':
+            _generated_text+=' '+_sentence+'.'
+        else:
+            _generated_text+=_sentence+'.'
+    return _generated_text
 
 def CreateFile(filename):
     __FILES.add(filename)
@@ -223,6 +232,8 @@ def Transcribe_V2(model, num_speakers, speaker_names, audio="temp_audio.wav"):
             #         conversation.append([GetSpeaker(segment["speaker"]), segment["text"][1:]]) # segment["speaker"] + ' ' + str(time(segment["start"])) + '\n\n'
             # conversation[-1][1] += segment["text"][1:]
         # return output
+        for idx in range(len(conversation)):
+            conversation[idx][3] = correct_grammar(conversation[idx][3])
         return ("".join([f"[{start}] - {speaker} \n{text}\n" for start, end, speaker, text in conversation])), ({ "data": [{"start": start, "end":end, "speaker": speaker, "text": text} for start, end, speaker, text in conversation]})
 
     def get_duration(path):