in decode_speech.py, line 242
pd.DataFrame(parsed_results).to_csv(f + '.txt', header=False, index=False)
saves out a blank text file.
The looks like the results from the google cloud speech-to-text are being parsed properly. Lines 154-57
for result in chunk.results:
alternative = result.alternatives[0]
print('Transcript: {}'.format(alternative.transcript))
print('Confidence: {}'.format(alternative.confidence))
print the transcripts and confidences properly and line 239
pickle.dump(results, open(f + ".p", "wb" ) )
saves out the pickled results object.
Could be an issue in pd.DataFrame.to_csv() inferring headers (header=Falsein line 242) or could be a suppressed error in line 229parsed_results = parse_response(results) when enable_word_time_offsets=False.
Will look into this later.
Code:
audiodir = os.path.abspath('../../data/audio/')
keypath = os.path.abspath('../../../google-credentials/cloud-speech-credentials.json')
for sid, data in rand5_turkids.items():
print(sid + ':')
for ses, turkid in data.items():
if ses == 'session 1':
audiofiles = [turkid+'-recall.wav', turkid+'-prediction.wav']
ep_context = ep1_speech_context
elif ses == 'session 2':
audiofiles = [turkid+'-delayed.wav', turkid+'-recall.wav']
if 'A' in sid:
ep_context = ep2_speech_context
elif 'B' in sid:
ep_context = ep3_speech_context
else:
print('bad SID: ' + sid)
continue
else:
print('session ID unrecognized for ' + sid + ', ' + ses)
continue
for audiofile in audiofiles:
path = os.path.join(audiodir,'room1',turkid, audiofile)
if not os.path.isfile(path):
path = os.path.join(audiodir,'room2',turkid, audiofile)
if not os.path.isfile(path):
print('AUDIO FILE NOT FOUND: ' + path)
continue
print('\tdecoding ' + ses + ': ' + audiofile + ' ...')
quail.decode_speech(path=path, keypath=keypath, save=True, speech_context=ep_context, enable_word_time_offsets=False)
in
decode_speech.py
, line 242pd.DataFrame(parsed_results).to_csv(f + '.txt', header=False, index=False)
saves out a blank text file.The looks like the results from the google cloud speech-to-text are being parsed properly. Lines 154-57
print the transcripts and confidences properly and line 239
pickle.dump(results, open(f + ".p", "wb" ) )
saves out the pickled results object.Could be an issue in
pd.DataFrame.to_csv() inferring headers (
header=Falsein line 242) or could be a suppressed error in line 229
parsed_results = parse_response(results) whenenable_word_time_offsets=False
. Will look into this later.Code: