malarinv
/
jasper-asr
mirror of https://github.com/malarinv/jasper-asr.git
1
0
Fork 0
Code Issues Releases Wiki Activity

respect verbose flag

Malar Kannan 2020-05-27 15:54:16 +05:30
parent 6f395af10d
commit 41af0a87de
2 changed files with 14 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
jasper/data/asr_recycler.py
View File

@ -59,11 +59,13 @@ def extract_data(
code_wav = code_fb.getvalue() code_wav = code_fb.getvalue()
# only starting 1 min audio has reliable alignment ignore rest # only starting 1 min audio has reliable alignment ignore rest
if start_time > 60: if start_time > 60:
print(f'start time over 60 seconds of audio skipping.') if verbose:
print(f'start time over 60 seconds of audio skipping.')
break break
# only if some reasonable audio data is present yield it # only if some reasonable audio data is present yield it
if code_seg.duration_seconds < 0.5: if code_seg.duration_seconds < 0.5:
print(f'transcript chunk "{transcript}" contains no audio skipping.') if verbose:
print(f'transcript chunk "{transcript}" contains no audio skipping.')
continue continue
yield transcript, code_seg.duration_seconds, code_wav yield transcript, code_seg.duration_seconds, code_wav

15
jasper/data/rev_recycler.py
View File

@ -63,7 +63,8 @@ def extract_data(
# print(monologue["speaker_name"]) # print(monologue["speaker_name"])
speaker_channel = channel_map.get(monologue["speaker_name"]) speaker_channel = channel_map.get(monologue["speaker_name"])
if not speaker_channel: if not speaker_channel:
print(f'unknown speaker tag {monologue["speaker_name"]} in wav:{wav_path} skipping.') if verbose:
print(f'unknown speaker tag {monologue["speaker_name"]} in wav:{wav_path} skipping.')
continue continue
try: try:
start_time = ( start_time = (
@ -79,7 +80,8 @@ def extract_data(
.collect()(monologue)[-1] .collect()(monologue)[-1]
) )
except IndexError: except IndexError:
print(f'error when loading timestamp events in wav:{wav_path} skipping.') if verbose:
print(f'error when loading timestamp events in wav:{wav_path} skipping.')
continue continue
# offset by 500 msec to include first vad? discarded audio # offset by 500 msec to include first vad? discarded audio
@ -92,7 +94,8 @@ def extract_data(
text_clean = re.sub(r"\[.*\]", "", text) text_clean = re.sub(r"\[.*\]", "", text)
# only if some reasonable audio data is present yield it # only if some reasonable audio data is present yield it
if tscript_wav_seg.duration_seconds < 0.5: if tscript_wav_seg.duration_seconds < 0.5:
print(f'transcript chunk "{text_clean}" contains no audio in {wav_path} skipping.') if verbose:
print(f'transcript chunk "{text_clean}" contains no audio in {wav_path} skipping.')
continue continue
yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav
@ -113,7 +116,8 @@ def extract_data(
.collect()(monologue)[-1] .collect()(monologue)[-1]
) )
except IndexError: except IndexError:
print(f'error when loading timestamp events in wav:{wav_path} skipping.') if verbose:
print(f'error when loading timestamp events in wav:{wav_path} skipping.')
continue continue
# offset by 500 msec to include first vad? discarded audio # offset by 500 msec to include first vad? discarded audio
@ -125,7 +129,8 @@ def extract_data(
text = "".join(lens["elements"].Each()["value"].collect()(monologue)) text = "".join(lens["elements"].Each()["value"].collect()(monologue))
text_clean = re.sub(r"\[.*\]", "", text) text_clean = re.sub(r"\[.*\]", "", text)
if tscript_wav_seg.duration_seconds < 0.5: if tscript_wav_seg.duration_seconds < 0.5:
print(f'transcript chunk "{text_clean}" contains no audio in {wav_path} skipping.') if verbose:
print(f'transcript chunk "{text_clean}" contains no audio in {wav_path} skipping.')
continue continue
yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav yield text_clean, tscript_wav_seg.duration_seconds, tscript_wav