-
Notifications
You must be signed in to change notification settings - Fork 8
/
preprocess_msra.py
89 lines (81 loc) · 3.55 KB
/
preprocess_msra.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
import re
import codecs
from pytorch_pretrained_bert.tokenization import BertTokenizer
tokenizer = BertTokenizer.from_pretrained('bert-pretrained/chinese_L-12_H-768_A-12', do_lower_case=True)
def preprocess(path, dataset):
label_type = ['O', 'NR', 'NS', 'NT']
end = [';', '?', '!', ';', '。', '?', '!']
comma = ','
max_length = 178
out_line_count = codecs.open('{}/{}_line_count.txt'.format(path, dataset), 'w', encoding='utf-8')
sens = []
for index, line in enumerate(codecs.open('{}/{}.txt'.format(path, dataset), encoding='utf-8')):
words, labels = [], []
for item in re.split(' +', line.strip()):
if item:
w = item.split('/')[0]
l = item.split('/')[1].upper()
if not w or l not in label_type:
print('{}: {}'.format(str(index), item))
continue
words.append(w)
labels.append(l)
assert len(words) == len(labels)
if len(''.join(words)) > (max_length - 20):
content_line, label_line, _split_result = [], [], []
i = 0
while i < len(words):
content_line.append(words[i])
label_line.append(labels[i])
if words[i] in end:
if i + 1 < len(words) and (words[i + 1] == '"' or words[i + 1] == '”'):
content_line.append(words[i + 1])
label_line.append(labels[i + 1])
i += 1
_split_result.append((content_line, label_line))
content_line, label_line = [], []
i += 1
if content_line:
_split_result.append((content_line, label_line))
split_result = []
for pair in _split_result:
if len(''.join(pair[0])) > (max_length - 20):
comma_index = [i for i, w in enumerate(pair[0]) if comma in w]
last_indx = 0
for indx in comma_index:
split_result.append((pair[0][last_indx: indx+1], pair[1][last_indx: indx+1]))
last_indx = indx + 1
if last_indx < len(pair[0]):
split_result.append((pair[0][last_indx:], pair[1][last_indx:]))
else:
split_result.append(pair)
else:
split_result = [(words, labels)]
out_line_count.write(str(len(split_result))+'\n')
sens.extend(split_result)
out_line_count.close()
out_content = codecs.open('{}/{}_content.txt'.format(path, dataset), 'w', encoding='utf-8')
out_label = codecs.open('{}/{}_label.txt'.format(path, dataset), 'w', encoding='utf-8')
for index, item in enumerate(sens):
content_line, label_line = item
units, labels = [], []
for word, label in zip(content_line, label_line):
_units = tokenizer.tokenize(word)
units.extend(_units)
if label == 'O':
start_l = 'O'
middle_l = 'O'
else:
start_l = 'B-' + label
middle_l = 'I-' + label
labels.append(start_l)
if len(_units) > 1:
for _ in _units[1:]:
labels.append(middle_l)
assert len(units) == len(labels), str(index)
if len(units) > max_length:
print(index)
out_content.write(' '.join(units)+'\n')
out_label.write(' '.join(labels)+'\n')
out_content.close()
out_label.close()