-
Notifications
You must be signed in to change notification settings - Fork 0
/
align_tdf_ph.py
212 lines (189 loc) · 7.79 KB
/
align_tdf_ph.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
""" Usage:
python3 align_tdf.py wavfile tdffile outputfile_alignment outputfile_words
"""
import re
import os
import sys
MODEL_DIR = '/app/aligner/english'
HVITE = '/app/htk/HTKTools/HVite'
HCOPY = '/app/htk/HTKTools/HCopy'
def prep_txt(trsfile, tmpbase, dictfile):
dict = []
with open(dictfile, 'r') as fid:
for line in fid:
dict.append(line.split()[0])
f = open(trsfile, 'r')
lines = f.readlines()
f.close()
fw = open(tmpbase + '.txt', 'w')
unk_words = []
first = True
tdf = False
for line in lines:
if first:
first = False
if re.match(r'^file;unicode', line):
tdf = True
continue
if tdf and re.match(r'^;;MM', line):
continue
line_split = line.split('\t')
if (len(line_split) >= 5):
if tdf:
st, en = line_split[2:4]
spkr = line_split[4]
txt = line_split[7]
else:
spkr, st, en = line_split[1:4]
txt = line_split[4]
for pun in ['*', '~', '--', ',', '.', ':', ';', '!', '?', '"', '(', ')', '+', '=']:
txt = txt.replace(pun, ' ')
wrds = []
for wrd in txt.split():
if (len(wrd) >= 2) and (wrd[-1] == '-'):
wrd = wrd[:-1]
if wrd[0] == "'":
wrd = wrd[1:]
if (wrd not in ["'", "-", ""]):
if (wrd.upper() not in dict):
unk_words.append(wrd.upper())
wrds.append('*' + wrd + '*')
else:
wrds.append(wrd)
if (len(wrds) > 0):
fw.write(st + '\t' + en + '\t' + spkr + '\t')
for wrd in wrds:
fw.write(wrd + ' ')
fw.write('\n')
fw.close()
#add unknown words to the standard dictionary, generate a tmp dictionary for alignment
fw = open(tmpbase + '.dict', 'w')
f = open(dictfile, 'r')
lines = f.readlines()
f.close()
for line in lines:
fw.write(line)
fw2 = open(tmpbase + '_unk.words', 'w')
for wrd in unk_words:
fw2.write(wrd + '\n')
if re.match(r'^\d+$', wrd):
fw.write('*' + wrd + '*\tAH1\n')
fw2.close()
fw3 = open(tmpbase + '_unk.words_lower', 'w')
for wrd in unk_words:
if wrd != '#':
fw3.write(wrd.lower() + '\n')
fw3.close()
os.system('phonetisaurus-apply --model /app/train/model.fst --word_list ' + tmpbase + '_unk.words_lower > ' + tmpbase + '_unk.words_with_phones')
f = open(tmpbase + '_unk.words_with_phones', 'r')
lines = f.readlines()
f.close()
for line in lines:
line_split = line.split('\t')
ww = line_split[0].upper()
newline = '*' + ww + '* ' + line_split[1]
if ww != '{LAUGH':
fw.write(newline)
fw.write('*{LAUGH*\tAH1\n')
fw.write('*}*\tAH1\n')
fw.write('*#*\tAH1\n')
fw.close()
def prep_mlf(txt, tmpbase):
fw = open(tmpbase + '.mlf', 'w')
fw.write('#!MLF!#\n')
fw.write('"' + tmpbase + '.lab"\n')
fw.write('sp\n')
for wrd in txt.split():
fw.write(wrd.upper() + '\n')
fw.write('sp\n')
fw.write('.\n')
fw.close()
def genres(tmpbase, alignfile, wordsfile):
f = open(tmpbase + '.txt', 'r')
lines = f.readlines()
f.close()
fw1 = open(alignfile, 'w')
fw2 = open(wordsfile, 'w')
fw1.write('#!MLF!#\n')
fw1.write('"' + tmpbase + '.rec"\n')
for i in range(len(lines)):
turn_st, turn_en, spkr, txt = lines[i].split('\t')
subtmpbase = tmpbase + '_' + str(i)
f = open(subtmpbase + '.aligned', 'r')
lls = f.readlines()
f.close()
if (len(lls) > 1):
times = []
j = 0
while (j < len(lls)):
if (j >= 2) and (j < (len(lls)-1)):
new_st= int(float(turn_st)*10000000) + int(lls[j].split()[0])
new_en = int(float(turn_st)*10000000) + int(lls[j].split()[1])
fw1.write(str(new_st) + ' ' + str(new_en))
for mm in lls[j].strip().split()[2:]:
fw1.write(' ' + mm)
fw1.write('\n')
if ((len(lls[j].split()) == 5) and (lls[j].split()[0] != lls[j].split()[1])):
wrd = lls[j].split()[-1].strip()
st = int(lls[j].split()[0])/10000000.0 + 0.0125 + float(turn_st)
k = j + 1
while (lls[k] != '.\n') and (len(lls[k].split()) != 5):
k += 1
en = int(lls[k-1].split()[1])/10000000.0 + 0.0125 + float(turn_st)
times.append([wrd, st, en])
j += 1
words = txt.strip().split()
words.reverse()
for item in times:
if (item[0] == 'sp'):
fw2.write(str(item[1]) + ' ' + str(item[2]) + ' ' + item[0] + ' ' + spkr + '\n')
else:
fw2.write(str(item[1]) + ' ' + str(item[2]) + ' ' + words.pop() + ' ' + spkr + '\n')
if (words != []):
print(lines[i], str(i) + '::not matched::' + alignfile)
else:
fw1.write(str(int(float(turn_st)*10000000)) + ' ' + str(int(float(turn_en)*10000000)) + ' ' + '***' + txt.strip().replace(' ', '_') + '***' + ' ' + '-1000000.0' + ' ' + '***' + txt.strip().replace(' ', '_') + '***' + '\n')
fw2.write(turn_st + ' ' + turn_en + ' ' + '***' + txt.strip().replace(' ', '_') + '***' + ' ' + spkr + '\n')
fw1.write('.\n')
fw1.close()
fw2.close()
if __name__ == '__main__':
try:
wavfile = sys.argv[1]
trsfile = sys.argv[2]
alignfile = sys.argv[3]
wordsfile = sys.argv[4]
except IndexError:
print("Input errors occurred!")
print(__doc__)
exit(1)
if 'USER' in os.environ:
tmpbase = './' + os.environ['USER'] + '_' + str(os.getpid())
else:
tmpbase = './' + os.environ['USERNAME'] + '_' + str(os.getpid())
samprate = os.popen('soxi -r ' + wavfile).read().strip()
#prepare clean_transcript file
prep_txt(trsfile, tmpbase, MODEL_DIR + '/dict')
f = open(tmpbase + '.txt', 'r')
lines = f.readlines()
f.close()
for i in range(len(lines)):
subtmpbase = tmpbase + '_' + str(i)
st = int(float(lines[i].split('\t')[0])*10000000)
en = int(float(lines[i].split('\t')[1])*10000000)
txt = lines[i].split('\t')[3]
prep_mlf(txt, subtmpbase)
#prepare scp
os.system(HCOPY + ' -C ' + MODEL_DIR + '/' + samprate + '/config ' + '-s ' + str(st) + ' ' + '-e ' + str(en) + ' ' + wavfile + ' ' + subtmpbase + '.plp')
#run alignment
if os.path.exists('/dev/null'):
os.system(HVITE + ' -a -m -t 10000.0 10000.0 100000.0 -I ' + subtmpbase + '.mlf -H ' + MODEL_DIR + '/' + samprate + '/macros -H ' + MODEL_DIR + '/' + samprate + '/hmmdefs -i ' + subtmpbase + '.aligned ' + tmpbase + '.dict ' + MODEL_DIR + '/monophones ' + subtmpbase + '.plp 2>&1 > /dev/null')
else:
os.system(HVITE + ' -a -m -t 10000.0 10000.0 100000.0 -I ' + subtmpbase + '.mlf -H ' + MODEL_DIR + '/' + samprate + '/macros -H ' + MODEL_DIR + '/' + samprate + '/hmmdefs -i ' + subtmpbase + '.aligned ' + tmpbase + '.dict ' + MODEL_DIR + '/monophones ' + subtmpbase + '.plp')
#generate results
genres(tmpbase, alignfile, wordsfile)
#clean up
print('Done!')
os.system('cp ' + tmpbase + '_unk.words' + ' ./' + trsfile.split('/')[-1].split('.')[0] + '.unks')
print('The unk words, whose pronunciations are generated by Phonetisaurus, are saved in ./' + trsfile.split('/')[-1].split('.')[0] + '.unks')
os.system('rm -f ' + tmpbase + '*')