if not leslie:
#make transcript file name
tf = dataset + '_transcript' + str(audio_file_num).zfill(3) + '.txt'
# print(lf)
#correct inconsistent naming
if dataset == 'tests':
inner_folder = 'test'
else:
inner_folder = dataset
#get sent timestamps (use memoisation to save too much disk access)
if TRANSCRIPT_DIR.format(inner_folder) + tf in memo:
sent_timestamps = memo[TRANSCRIPT_DIR.format(inner_folder) + tf]
else:
_, sent_timestamps, _ = process_transcript(
TRANSCRIPT_DIR.format(inner_folder) + tf)
memo[TRANSCRIPT_DIR.format(inner_folder) + tf] = sent_timestamps
#get start timestamp for this utterance (in seconds)
#note we probs won't need end_time
#(but can use it to sanity check that we have the correct file if u want...)
utt_start, _ = sent_timestamps[sent_num - 1]
# print(sent_num-1, sent_timestamps, utt_start, '\n')
#use dict to keep the data for this dialog file for each emotion
#key='arousal' value=[0.3004, 0.3005, 0.3006, ...]
emotion_data = {}
#load emotion data in for this dialog
for emotion in ['arousal', 'expectancy', 'power', 'valence']:
#get the continuous emotion feature file paths
# print(transcript_files, wav_files)
#work through each dialog one at a time
for transcript_file, wav_file in zip(transcript_files, wav_files):
if INCLUDE_LIST:
if transcript_file not in INCLUDE_LIST:
print(transcript_file, "NOT IN INCLUDE LIST")
continue #will skip over any files which aren't in the include list
if EXCLUDE_LIST:
if transcript_file in EXCLUDE_LIST:
print(transcript_file, "IN EXCLUDE LIST")
continue
#read in transcript file and get sentences, sent and word time stamps
output_sentences, sent_timestamps, word_timestamps = process_transcript(INPUT_FOLDER + TRANSCRIPT_FOLDER + inner_folder + transcript_file)
#use sentence timestamps to segment the dialog wav files into wav files only containing sentences
#also create txt files that contain transcriptions for each sentence
i = 1
for sentence, timestamps in zip(output_sentences, sent_timestamps):
begin = str(datetime.timedelta(seconds=timestamps[0]))
end = str(datetime.timedelta(seconds=timestamps[1]))
file_name = wav_file[:-4] #get rid of .wav extension
file_name = file_name + '_sent{}'.format(str(i).zfill(3))
#if a filename is in the exclude sent list, then continue this for loop
if file_name in SENT_EXCLUDE_LIST:
print(file_name, 'in SENT_EXCLUDE_LIST')
else: