-
Notifications
You must be signed in to change notification settings - Fork 0
/
preprocess.py
183 lines (157 loc) · 5.34 KB
/
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
import os
import json
import pandas as pd
import re
from random import shuffle, choice
from util import load_pair, load_poly
path_zh_en = 'dict/zh_en.csv'
path_homo = 'dict/homo.csv'
path_syno = 'dict/syno.csv'
zh_en = load_pair(path_zh_en)
homo_dict = load_poly(path_homo)
syno_dict = load_poly(path_syno)
def save(path, sents):
with open(path, 'w') as f:
json.dump(sents, f, ensure_ascii=False, indent=4)
def dict2list(sents):
word_mat, label_mat = list(), list()
for pairs in sents.values():
words, labels = list(), list()
for pair in pairs:
words.append(pair['word'])
labels.append(pair['label'])
word_mat.append(words)
label_mat.append(labels)
return word_mat, label_mat
def list2dict(word_mat, label_mat):
sents = dict()
for words, labels in zip(word_mat, label_mat):
text = ''.join(words)
pairs = list()
for word, label in zip(words, labels):
pair = dict()
pair['word'] = word
pair['label'] = label
pairs.append(pair)
sents[text] = pairs
return sents
def select(part):
if part[0] == '[' and part[-1] == ']':
word = part[1:-1]
cands = set()
cands.add(word)
if word in syno_dict:
cands.update(syno_dict[word])
if word in homo_dict:
cands.update(homo_dict[word])
return choice(list(cands))
elif part[0] == '(' and part[-1] == ')':
word = part[1:-1]
return choice([word, ''])
else:
return part
def generate(temps, slots, num):
word_mat, label_mat = list(), list()
for i in range(num):
parts = choice(temps)
words, labels = list(), list()
for part in parts:
if part in slots:
entity = choice(slots[part])
words.extend(entity)
labels.append('B-' + part)
if len(entity) > 1:
labels.extend(['I-' + part] * (len(entity) - 1))
else:
word = select(part)
if word:
words.extend(word)
labels.extend(['O'] * len(word))
word_mat.append(words)
label_mat.append(labels)
return word_mat, label_mat
def label_sent(path):
sents = dict()
for text, entity_str, label_str in pd.read_csv(path).values:
entitys, labels = entity_str.split(), label_str.split()
if len(entitys) != len(labels):
print('skip: %s', text)
continue
slots = ['O'] * len(text)
for entity, label in zip(entitys, labels):
heads = [iter.start() for iter in re.finditer(entity, text)]
entity_len = len(entity)
for head in heads:
tail = head + entity_len
if slots[head:tail] != ['O'] * entity_len:
print('skip: %s in %s' % (entity, text))
continue
slots[head] = 'B-' + label
for i in range(1, entity_len):
slots[head + i] = 'I-' + label
pairs = list()
for word, label in zip(text, slots):
pair = dict()
pair['word'] = word
pair['label'] = label
pairs.append(pair)
sents[text] = pairs
return sents
def merge_sent(path):
sents = dict()
pairs = list()
with open(path, 'r') as f:
for line in f:
line = line.strip()
if line:
pair = dict()
word, label = line.split()
pair['word'] = word
pair['label'] = label
pairs.append(pair)
elif pairs:
text = ''.join([pair['word'] for pair in pairs])
sents[text] = pairs
pairs = []
return sents
def expand(sents, gen_word_mat, gen_label_mat):
word_mat, label_mat = dict2list(sents)
word_mat.extend(gen_word_mat)
label_mat.extend(gen_label_mat)
pairs = list(zip(word_mat, label_mat))
shuffle(pairs)
word_mat, label_mat = zip(*pairs)
bound = int(len(word_mat) * 0.9)
train_sents = list2dict(word_mat[:bound], label_mat[:bound])
test_sents = list2dict(word_mat[bound:], label_mat[bound:])
return train_sents, test_sents
def prepare(paths):
temps = list()
with open(paths['temp'], 'r') as f:
for line in f:
parts = line.strip().split()
temps.append(parts)
slots = dict()
files = os.listdir(paths['slot_dir'])
for file in files:
label = zh_en[os.path.splitext(file)[0]]
slots[label] = list()
with open(os.path.join(paths['slot_dir'], file), 'r') as f:
for line in f:
slots[label].append(line.strip())
gen_word_mat, gen_label_mat = generate(temps, slots, num=5000)
sent1s = merge_sent(paths['univ'])
sent2s = label_sent(paths['extra'])
sents = dict(sent1s, **sent2s)
train_sents, test_sents = expand(sents, gen_word_mat, gen_label_mat)
save(paths['train'], train_sents)
save(paths['test'], test_sents)
if __name__ == '__main__':
paths = dict()
paths['univ'] = 'data/univ.txt'
paths['train'] = 'data/train.json'
paths['test'] = 'data/test.json'
paths['temp'] = 'data/template.txt'
paths['slot_dir'] = 'data/slot'
paths['extra'] = 'data/extra.csv'
prepare(paths)