forked from nguyenvulebinh/lyric-alignment
-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
218 lines (187 loc) · 7.28 KB
/
utils.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
215
216
217
218
import json
from vinorm import TTSnorm
from regtag import augment, reoov, clean_text
import re
from dataclasses import dataclass
import torch
import glob
import os
import shutil
with open('./read_map.json', 'r', encoding='utf-8') as file:
exeption_oov = json.load(file)
@dataclass
class Point:
token_index: int
time_index: int
score: float
# Merge the labels
@dataclass
class Segment:
label: str
start: int
end: int
score: float
def __repr__(self):
return f"{self.label}\t({self.score:4.2f}): [{self.start:5d}, {self.end:5d})"
@property
def length(self):
return self.end - self.start
def itn_text(word):
word = word.lower().strip(' -.?!,)("\'…“”*_[]’')
word = re.sub(r'[ {}]'.format(re.escape('-.?!,)("\'…“”*_[]’')), ' ', word)
word = re.sub(r'\s+', ' ', word)
if reoov.check_oov_word(word) or reoov.format_word(word) not in reoov.vi_dict or augment.get_random_oov(word) is not None or exeption_oov.get(word, None) is not None:
if exeption_oov.get(word, None) is not None:
return exeption_oov[word]
read_form = augment.get_random_oov(word)
if read_form is not None:
tgt = augment.oov_dict[word]
tgt.sort()
tgt = tgt[0]
return tgt
if len(set(list(word)).intersection(set(list('0123456789')))) > 0:
read_form = TTSnorm(word).lower().strip('. ')
return read_form
return word
def norm_word(text):
text = re.sub('\s+', ' ', text).strip(' -.?!,)("\'…“”*_[]’')
words = text.lower().strip('., :?').split()
norm = [itn_text(w) for w in words]
return ' '.join(norm)
def get_trellis(emission, tokens, blank_id=0):
num_frame = emission.size(0)
num_tokens = len(tokens)
# Trellis has extra diemsions for both time axis and tokens.
# The extra dim for tokens represents <SoS> (start-of-sentence)
# The extra dim for time axis is for simplification of the code.
trellis = torch.empty((num_frame + 1, num_tokens + 1))
trellis[0, 0] = 0
trellis[1:, 0] = torch.cumsum(emission[:, 0], 0)
trellis[0, -num_tokens:] = -float("inf")
trellis[-num_tokens:, 0] = float("inf")
for t in range(num_frame):
trellis[t + 1, 1:] = torch.maximum(
# Score for staying at the same token
trellis[t, 1:] + emission[t, blank_id],
# Score for changing to the next token
trellis[t, :-1] + emission[t, tokens],
)
return trellis
def backtrack(trellis, emission, tokens, blank_id=0):
# Note:
# j and t are indices for trellis, which has extra dimensions
# for time and tokens at the beginning.
# When referring to time frame index `T` in trellis,
# the corresponding index in emission is `T-1`.
# Similarly, when referring to token index `J` in trellis,
# the corresponding index in transcript is `J-1`.
j = trellis.size(1) - 1
t_start = torch.argmax(trellis[:, j]).item()
path = []
for t in range(t_start, 0, -1):
# 1. Figure out if the current position was stay or change
# Note (again):
# `emission[J-1]` is the emission at time frame `J` of trellis dimension.
# Score for token staying the same from time frame J-1 to T.
stayed = trellis[t - 1, j] + emission[t - 1, blank_id]
# Score for token changing from C-1 at T-1 to J at T.
changed = trellis[t - 1, j - 1] + emission[t - 1, tokens[j - 1]]
# 2. Store the path with frame-wise probability.
prob = emission[t - 1, tokens[j - 1] if changed > stayed else 0].exp().item()
# Return token index and time index in non-trellis coordinate.
path.append(Point(j - 1, t - 1, prob))
# 3. Update the token
if changed > stayed:
j -= 1
if j == 0:
break
else:
raise ValueError("Failed to align")
return path[::-1]
def merge_repeats(path, transcript):
i1, i2 = 0, 0
segments = []
while i1 < len(path):
while i2 < len(path) and path[i1].token_index == path[i2].token_index:
i2 += 1
score = sum(path[k].score for k in range(i1, i2)) / (i2 - i1)
segments.append(
Segment(
transcript[path[i1].token_index],
path[i1].time_index,
path[i2 - 1].time_index + 1,
score,
)
)
i1 = i2
return segments
# Merge words
def merge_words(segments, separator="|"):
words = []
i1, i2 = 0, 0
while i1 < len(segments):
if i2 >= len(segments) or segments[i2].label == separator:
if i1 != i2:
segs = segments[i1:i2]
word = "".join([seg.label for seg in segs])
score = sum(seg.score * seg.length for seg in segs) / sum(seg.length for seg in segs)
words.append(Segment(word, segments[i1].start, segments[i2 - 1].end, score))
i1 = i2 + 1
i2 = i1
else:
i2 += 1
return words
def add_pad(word_segments, emission, shift_val=1):
for i in range(len(word_segments) - 1):
word = word_segments[i]
next_word = word_segments[i + 1]
next_word.start -= shift_val
if next_word.start - word.start > 150:
word.end = word.start + 150
else:
word.end = next_word.start
next_word.start -= shift_val
# if word.end - word.start <= 70:
# word.start -= 1
# word.end += 1
# elif word.end - word.start < 7:
# word.start -= 2
# word.end += 2
word = word_segments[-1]
word_segments[-1].end = min(len(emission), word_segments[-1].end + 200)
word_segments[0].start = max(0, word_segments[0].start - 50)
return word_segments
def shift_align(lyric_alignment, shift_ms=120):
words = [item for sublist in [seg['l'] for seg in lyric_alignment] for item in sublist]
for i in range(len(words)):
# print(words[i])
shif_val_s = 0 if (words[i]['s'] - shift_ms) < 0 else shift_ms
shif_val_e = 0 if (words[i]['e'] - shift_ms) < words[i]['s'] else shift_ms
words[i]['s'] -= shif_val_s
words[i]['e'] -= shif_val_e
# print(shif_val, words[i], '\n')
for seg in lyric_alignment:
if len(seg['l']) > 0:
seg['s'] = seg['l'][0]['s']
seg['e'] = seg['l'][-1]['e']
return lyric_alignment
def load_test_case(data_path):
all_wav_file = glob.glob(os.path.join(data_path, '**/*_normed_16k.wav'), recursive=True)
all_json_file = glob.glob(os.path.join(data_path, '**/*.json'), recursive=True)
map_test_case = dict({})
for item in all_wav_file:
id = item.split('/')[-1].split('_')[0]
if map_test_case.get(id, None) is None:
map_test_case[id] = {
'path_wav': item
}
for item in all_json_file:
id = item.split('/')[-1].split('.')[0]
if map_test_case.get(id, None) is not None:
map_test_case[id]['path_lyric'] = item
return map_test_case
def zip_folder(input_folder_path, output_zip_path):
shutil.make_archive(output_zip_path, 'zip', input_folder_path)
if __name__ == "__main__":
# print(itn_text('world....'))
print(load_test_case('./data'))