bug fix for transcription

Author: nonstoptimm

src/glue.py

@@ -76,7 +76,7 @@
             logging.info('[INFO] - Starting with speech-to-text conversion')
             stt_results = stt.main(f'{audio_files}/', f'{output_folder}/{case}')
             df_transcription = pd.DataFrame(list(stt_results), columns=['audio', 'rec'])
-            logging.debug(transcription)
+            logging.debug(df_transcription)
             df_transcription.to_csv(f'{output_folder}/{case}/stt_transcriptions.txt', sep = '\t', header = None, index=False)
             # Merge reference transcriptions with recognition on audio file names
             if 'audio' in list(df_reference.columns):

src/tts.py

@@ -181,8 +181,9 @@ def main(df, output_directory, custom=True, telephone=True):
             fname = "nan"
         audio_synth.append(fname)
     df['audio_synth'] = audio_synth
-    df['text_clean'] = df.text.apply(remove_tags)
+    df['text_ssml'] = df['text'].copy()
+    df['text'] = df['text_ssml'].apply(remove_tags)
     return df
 
 if __name__ == '__main__':
-    main(pd.DataFrame({'text': ['Ich möchte diesen Teppicht nicht kaufen', 'Was geht los da rein?']}), "output/test")
+    main(pd.DataFrame({'text': ['This is a test', 'And this is another test!']}), "output/test")