forked from gooofy/zamia-speech
-
Notifications
You must be signed in to change notification settings - Fork 2
/
import_trs.py
executable file
·214 lines (187 loc) · 9.87 KB
/
import_trs.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
#!/usr/bin/env python
# -*- coding: utf-8 -*-
# Copyright 2019 by Paul Guyot
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# convert Transcriber files to voxforge-style packages
#
# Import .trs and .wav/.mp3 files. File names should match. A heuristic is used
# when they do not match exactly, however it may fail or process the wrong file.
# Run with -n to check all files can be read and they match.
# Volume is adjusted on a per-file basis.
# Turns with no speaker attribute or with several speakers are ignored.
import sys
import os
import codecs
import traceback
import logging
import re
import xml.etree.ElementTree as ET
import multiprocessing as mp
from difflib import SequenceMatcher
from optparse import OptionParser
from nltools import misc
PROC_TITLE = 'trs_to_vf'
#
# init terminal
#
misc.init_app (PROC_TITLE)
#
# command line
#
parser = OptionParser("usage: %prog [options] corpus")
parser.add_option ("-v", "--verbose", action="store_true", dest="verbose",
help="enable debug output")
parser.add_option ("-n", "--dry", action="store_true", dest="dry",
help="check files, do not actually generate files")
(options, args) = parser.parse_args()
if options.verbose:
logging.basicConfig(level=logging.DEBUG)
else:
logging.basicConfig(level=logging.INFO)
if len(args)<1:
parser.print_usage()
sys.exit(1)
corpus_name = args[0]
#
# config
#
config = misc.load_config ('.speechrc')
speech_arc = config.get("speech", "speech_arc")
speech_corpora = config.get("speech", "speech_corpora")
# iterate on every trs file in arc/corpus_name/
class DoctypeAwareTreeBuilder(ET.TreeBuilder):
def doctype(self, name, pubid, system):
self.__doctype_name = name
self.__doctype_system = system
def valid(self):
return self.__doctype_name == 'Trans' and (self.__doctype_system == 'trans-14.dtd' or self.__doctype_system == 'trans-13.dtd')
def seconds_to_ffmpeg_ts(seconds):
minutes=int(seconds/60)
seconds=seconds-minutes*60
hours=int(minutes/60)
minutes=minutes-hours*60
return '{:02d}:{:02d}:{:06.3f}'.format(hours, minutes, seconds)
def get_volume(recording):
cmd="ffmpeg -hide_banner -i {recording} -filter:a volumedetect -f null /dev/null 2>&1 | grep max_volume | awk '{{ print $5 }}'".format(recording=recording)
volume=os.popen(cmd).read()
volume=-float(volume)
return volume
def extract_wav(recording, volume, wav, startTime, endTime):
if os.path.exists(wav):
logging.debug('Skipping {wav} as it already exists'.format(wav=wav))
return
startTimeF = float(startTime)
endTimeF = float(endTime)
duration = endTimeF - startTimeF
startTimeFormatted=seconds_to_ffmpeg_ts(startTimeF)
durationFormatted=seconds_to_ffmpeg_ts(duration)
cmd='ffmpeg -y -hide_banner -loglevel panic -i {recording} -ss {startTimeFormatted} -t {durationFormatted} -filter:a "volume={volume}dB" -ac 1 -ar 16k {wav}'.format(recording=recording, startTimeFormatted=startTimeFormatted, durationFormatted=durationFormatted, volume=volume, wav=wav)
logging.debug('{cmd}'.format(cmd=cmd))
os.system(cmd)
def find_recording(dir, root_filename):
for suffix in ['.wav', '.WAV', '.mp3', '.mp4', '-v2.wav', '-v2.mp3']:
recording='{dir}/{root_filename}{suffix}'.format(dir=dir, root_filename=root_filename, suffix=suffix)
if os.path.isfile(recording):
return recording
best_candidate=None
best_candidate_ratio=0.9
for recording_filename in os.listdir(dir):
if recording_filename.endswith(".mp3") or recording_filename.endswith(".wav"):
basename=recording_filename[:-4]
candidate_ratio = SequenceMatcher(None, basename, root_filename).ratio()
if candidate_ratio > best_candidate_ratio:
best_candidate=recording_filename
best_candidate_ratio=candidate_ratio
if best_candidate!=None:
logging.warning('{root_filename}: using file {recording}, ratio={ratio}'.format(root_filename=root_filename, recording=best_candidate, ratio=best_candidate_ratio))
return '{dir}/{best_candidate}'.format(dir=dir, best_candidate=best_candidate)
logging.error('{root_filename}: cannot find recording file'.format(root_filename=root_filename))
return None
def process_dir(pool, dir):
totalDuration=0
for filename in os.listdir(dir):
path='%s/%s' % (dir, filename)
if os.path.isdir(path):
totalDuration=totalDuration+process_dir(pool, path)
elif os.path.isfile(path) and filename.endswith(".trs"):
corpus_filename=filename
corpus_filepath=path
logging.info('Processing {corpus_filename}'.format(corpus_filename=corpus_filename))
root_filename=corpus_filename[:-4]
if root_filename.endswith("-v2"):
root_filename=root_filename[:-3]
recording=find_recording(dir, root_filename)
if recording == None:
continue
if options.dry and not options.verbose:
volume=None
else:
volume=get_volume(recording)
treebuilder = DoctypeAwareTreeBuilder()
parser = ET.XMLParser(target=treebuilder)
tree = ET.parse(corpus_filepath, parser=parser)
if not treebuilder.valid():
logging.error('{corpus_filepath}: Invalid DOCTYPE, was expecting <!DOCTYPE Trans SYSTEM "trans-13.dtd"> or <!DOCTYPE Trans SYSTEM "trans-14.dtd">'.format(corpus_filepath=corpus_filepath))
continue
root = tree.getroot()
speakerst = {}
for speakers in root.findall('Speakers'):
for speaker in speakers.findall('Speaker'):
speakerid = speaker.attrib['id']
speakername = speaker.attrib['name']
speakerst[speakerid] = re.sub(r'[ ()/-]', '', speakername)
for episode in root.findall('Episode'):
for section in episode.findall('Section'):
for turn in section.findall('Turn'):
if not 'speaker' in turn.attrib:
logging.warning('{corpus_filename}: Turn element missing speaker attribute'.format(corpus_filename=corpus_filename))
continue
speakerid = turn.attrib['speaker']
if speakerid in speakerst: # skip cases with several speakers
startTime = turn.attrib['startTime']
endTime = turn.attrib['endTime']
speakername = speakerst[speakerid]
startTimeF = float(startTime)
endTimeF = float(endTime)
duration = endTimeF - startTimeF
wavdir='{corpora}/{corpus_name}/{spk}-{name}/wav/'.format(corpora=speech_corpora, corpus_name=corpus_name, name=root_filename, spk=speakername)
etcdir='{corpora}/{corpus_name}/{spk}-{name}/etc/'.format(corpora=speech_corpora, corpus_name=corpus_name, name=root_filename, spk=speakername)
if not options.dry:
if not os.path.exists(wavdir):
os.makedirs(wavdir)
if not os.path.exists(etcdir):
os.makedirs(etcdir)
text=" ".join(turn.itertext()).replace('\n', '').replace(' ', ' ')
if re.search(r'\w', text, re.I) != None and re.search(r'¤', text) == None:
uttid='{spk}-{name}-{startTime}-{endTime}'.format(name=root_filename, spk=speakername, startTime=startTime, endTime=endTime)
if not options.dry:
with codecs.open('{etcdir}/prompts-original'.format(etcdir=etcdir), 'a', 'utf8') as promptsf:
promptsf.write('%s %s\n' % (uttid, text))
wav='{corpora}/{corpus_name}/{spk}-{name}/wav/{uttid}.wav'.format(corpora=speech_corpora, corpus_name=corpus_name, name=root_filename, spk=speakername, uttid=uttid)
if options.dry:
logging.debug('Would convert {recording} from {startTime} to {endTime} to {wav} with volume gain of {volume}'.format(recording=recording, volume=volume, wav=wav, startTime=startTime, endTime=endTime))
else:
logging.debug('Convert {recording} from {startTime} to {endTime} to {wav} with volume gain of {volume}'.format(recording=recording, volume=volume, wav=wav, startTime=startTime, endTime=endTime))
pool.apply_async(extract_wav, args=(recording, volume, wav, startTime, endTime))
totalDuration=totalDuration+duration
else:
logging.debug('Skipping empty/anonymized text={text}'.format(text=text))
return totalDuration
pool = mp.Pool(mp.cpu_count())
arc_corpus_dir='{speech_arc}/{corpus_name}'.format(speech_arc=speech_arc, corpus_name=corpus_name)
totalDuration=process_dir(pool, arc_corpus_dir)
logging.info('Total duration: {formatted}'.format(formatted=seconds_to_ffmpeg_ts(totalDuration)))
pool.close()
pool.join()