-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathdata_preprocess.py
213 lines (172 loc) · 7.46 KB
/
data_preprocess.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
import argparse
import pickle
from tqdm import tqdm
from collections import defaultdict
from transformers import BertTokenizer
from tokenizers import normalizers
from tokenizers.normalizers import NFD, NFKD, NFC, NFKC, Lowercase, StripAccents, Strip, Replace, BertNormalizer
def construct_parallel_data_para(src_path, trg_path):
parallel_data = []
with open(src_path, "r") as f:
src_lines = f.readlines()
with open(trg_path, "r") as f:
trg_lines = f.readlines()
assert len(src_lines) == len(trg_lines)
print("data size: " + str(len(src_lines)))
c_no_error_sent = 0
for src_line, trg_line in zip(src_lines, trg_lines):
src_items = src_line.strip().split("\t")
assert len(src_items) == 2
src_sent = src_items[1]
trg_items = trg_line.strip().split("\t")
assert trg_items[0] == src_items[0]
id = trg_items[0]
trg_sent = trg_items[1]
modification = []
if src_sent == trg_sent:
c_no_error_sent += 1
else:
for i, (src_char, trg_char) in enumerate(zip(src_sent, trg_sent)):
if src_char != trg_char:
modification.append((i, trg_char))
parallel_data.append((id, src_sent, trg_sent, modification))
print("error-free sentences: " + str(c_no_error_sent))
return parallel_data
def encode_parallel_data(config, parallel_data, normalizer, tokenizer, max_len, ids2pinyin):
data = []
for item in tqdm(parallel_data):
data_sample = {}
if config.normalize == "True":
src_norm = normalizer.normalize_str(item[1])[:max_len - 2]
trg_norm = normalizer.normalize_str(item[2])[:max_len - 2]
else:
src_norm = item[1][:max_len - 2]
trg_norm = item[2][:max_len - 2]
if len(src_norm) != len(trg_norm):
pass
src_token_list = list(src_norm)
trg_token_list = list(trg_norm)
src_token_list.insert(0, '[CLS]')
src_token_list.append('[SEP]')
trg_token_list.insert(0, '[CLS]')
trg_token_list.append('[SEP]')
data_sample['id'] = item[0]
data_sample['src_text'] = item[1]
data_sample['input_ids'] = tokenizer.convert_tokens_to_ids(src_token_list)
data_sample['token_type_ids'] = [0 for _ in range(len(src_token_list))]
data_sample['attention_mask'] = [1 for _ in range(len(src_token_list))]
data_sample['pinyin_ids'] = [ids2pinyin[i] for i in data_sample['input_ids']]
data_sample['trg_ids'] = tokenizer.convert_tokens_to_ids(trg_token_list)
data_sample['trg_text'] = item[2]
data_sample['modification'] = item[3]
data.append(data_sample)
return data
def construct_parallel_data_lbl(src_path, trg_path):
parallel_data = []
with open(src_path, "r") as f:
src_lines = f.readlines()
with open(trg_path, "r") as f:
trg_lines = f.readlines()
assert len(src_lines) == len(trg_lines)
print("data size: " + str(len(src_lines)))
c_no_error_sent = 0
for src_line, trg_line in zip(src_lines, trg_lines):
src_items = src_line.strip().split("\t")
assert len(src_items) == 2
src_sent = src_items[1]
trg_items = trg_line.strip().split(", ")
id = trg_items[0]
trg_sent = list(src_sent)
modification = []
if len(trg_items) == 2:
c_no_error_sent += 1
else:
for i in range(1, len(trg_items), 2):
trg_sent[int(trg_items[i]) - 1] = trg_items[i + 1]
modification.append((int(trg_items[i]) - 1, trg_items[i + 1]))
trg_sent = "".join(trg_sent)
parallel_data.append((id, src_sent, trg_sent, modification))
print("error-free sentences: " + str(c_no_error_sent))
return parallel_data
def encode_predict_data(config, src_path, normalizer, tokenizer, max_len, ids2pinyin):
data = []
with open(src_path, "r") as f:
src_lines = f.readlines()
print("data size: " + str(len(src_lines)))
for src_line in src_lines:
data_sample = {}
src_items = src_line.strip().split("\t")
assert len(src_items) == 2
src_sent = src_items[1]
id = src_items[0]
if config.normalize == "True":
src_norm = normalizer.normalize_str(src_sent)[:max_len - 2]
else:
src_norm = normalizer.normalize_str(src_sent)[:max_len - 2]
src_token_list = list(src_norm)
src_token_list.insert(0, '[CLS]')
src_token_list.append('[SEP]')
data_sample['id'] = id
data_sample['src_text'] = src_sent
data_sample['input_ids'] = tokenizer.convert_tokens_to_ids(src_token_list)
data_sample['token_type_ids'] = [0 for i in range(len(src_token_list))]
data_sample['attention_mask'] = [1 for i in range(len(src_token_list))]
data_sample['pinyin_ids'] = [ids2pinyin[i] for i in data_sample['input_ids']]
data.append(data_sample)
return data
def save_as_pkl(data, path):
with open(path, 'wb') as f:
pickle.dump(data, f)
def main(config):
print(config.__dict__)
tokenizer = BertTokenizer.from_pretrained(config.bert_path)
ids2pinyin = defaultdict(int)
with open("pymodel/vocab/pinyin_mapping.txt") as f:
for l in f:
ids, pinyin = l.strip().split()
ids2pinyin[int(ids)] = int(pinyin)
do_zimu = False
if do_zimu:
ZM2ID = {':': 1, 'a': 2, 'c': 3, 'b': 4, 'e': 5, 'd': 6, 'g': 7, 'f': 8, 'i': 9, 'h': 10, 'k': 11, 'j': 12,
'm': 13, 'l': 14, 'o': 15, 'n': 16, 'q': 17, 'p': 18, 's': 19, 'r': 20, 'u': 21, 't': 22, 'w': 23,
'v': 24, 'y': 25, 'x': 26, 'z': 27}
PYLEN = 5
ids2zimu = defaultdict(int)
f = open('pymodel/vocab/pinyin_vocab.txt')
lines = f.readlines()
for k in ids2pinyin:
pinyin = lines[ids2pinyin[k]].strip()
if pinyin != '[OTHER]':
seq = []
for c in pinyin:
seq.append(ZM2ID[c])
seq = [0] * PYLEN + seq
seq = seq[-PYLEN:]
ids2zimu[k] = seq
else:
ids2zimu[k] = [0] * PYLEN
normalizer = normalizers.Sequence([Lowercase()])
if config.target_dir:
if config.data_mode == "para":
parallel_data = construct_parallel_data_para(config.source_dir, config.target_dir)
elif config.data_mode == "lbl":
parallel_data = construct_parallel_data_lbl(config.source_dir, config.target_dir)
else:
print("Wrong data mode!")
exit()
encode_data = encode_parallel_data(config, parallel_data, normalizer, tokenizer, config.max_len, ids2pinyin)
save_as_pkl(encode_data, config.save_path)
else:
encode_data = encode_predict_data(config, config.source_dir, normalizer, tokenizer, config.max_len, ids2pinyin)
save_as_pkl(encode_data, config.save_path)
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--source_dir", required=True, type=str)
parser.add_argument("--target_dir", type=str)
parser.add_argument("--bert_path", default="bert-base-chinese", type=str)
parser.add_argument("--max_len", default=128, type=int)
parser.add_argument("--save_path", required=True, type=str)
parser.add_argument("--data_mode", required=True, type=str)
parser.add_argument("--normalize", default="True", type=str)
args = parser.parse_args()
main(args)