-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathgenerate_ner.py
222 lines (188 loc) · 9.54 KB
/
generate_ner.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
219
220
221
222
import spacy
import os
from data_utils import load_sentences4spacy, load_labels
base_dir = os.path.dirname(__file__)
# The path of the dataset is modified in there
# 获取当前文件目录,替换middle参数即可。
# new_middle = 'new_atis_part'
# middle = 'atis'
# mid_middle = 'tmp_atis'
# new_middle = 'new_atis_all'
# middle = 'atis'
# mid_middle = 'tmp_atis'
# new_middle = 'new_snips_all'
# middle = 'snips'
# mid_middle = 'tmp_snips'
# new_middle = 'new_snips_part'
# middle = 'snips'
# mid_middle = 'tmp_snips'
# new_middle = 'new_slur_all'
# middle = 'slur'
# mid_middle = 'tmp_slur'
new_middle = 'new_slur_part'
middle = 'slur'
mid_middle = 'tmp_slur'
train_ner_path = os.path.join(base_dir, middle, 'train/seq.ner')
new_train_in_path = os.path.join(base_dir, new_middle, 'train/seq.in')
new_train_ner_path = os.path.join(base_dir, new_middle, 'train/seq.ner')
new_train_out_path = os.path.join(base_dir, new_middle, 'train/seq.out')
new_train_label_path = os.path.join(base_dir, new_middle, 'train/label')
valid_ner_path = os.path.join(base_dir, middle, 'valid/seq.ner')
new_valid_in_path = os.path.join(base_dir, new_middle, 'valid/seq.in')
new_valid_ner_path = os.path.join(base_dir, new_middle, 'valid/seq.ner')
new_valid_out_path = os.path.join(base_dir, new_middle, 'valid/seq.out')
new_valid_label_path = os.path.join(base_dir, new_middle, 'valid/label')
test_ner_path = os.path.join(base_dir, middle, 'test/seq.ner')
new_test_in_path = os.path.join(base_dir, new_middle, 'test/seq.in')
new_test_ner_path = os.path.join(base_dir, new_middle, 'test/seq.ner')
new_test_out_path = os.path.join(base_dir, new_middle, 'test/seq.out')
new_test_label_path = os.path.join(base_dir, new_middle, 'test/label')
# Load English tokenizer, tagger, parser, NER and word vectors2
# nlp = spacy.load("en_core_web_sm")
nlp = spacy.load("en_core_web_lg")
def ner_text(sentence_list, new_ner_path): # 去掉空NER行，做数据的筛选，生成新的数据集
j = 0 # j 用来统计全是空NER标签的句子数量，进而计算比例
need_list = []
with open(new_ner_path, 'w') as seqner:
for i, sentence in enumerate(sentence_list):
sentence_str = ' '.join(sentence)
doc = nlp(sentence_str)
dic_ner = {}
for entity in doc.ents:
# print(entity.text, ':', entity.label_)
# print(len(entity.text.split(' ')))
# 一个单词的情况
if len(entity.text.split(' ')) == 1:
dic_ner[entity.text] = entity.label_
# 对于一个单词的情况
if len(entity.text.split(' ')) > 1:
for word in entity.text.split(' '):
if word == entity.text.split(' ')[0]:
dic_ner[word] = entity.label_
else:
dic_ner[word] = entity.label_
if doc.ents: # 如果这行数据有NER标签,则写入数据，否则这条数据忽略
# for entity in doc.ents:
# print(entity.text,entity.label_)
j = j + 1
need_list.append(i) # 有NER数据的时候就把这个数据的ID保存下来
for position, word in enumerate(sentence):
# print(position, len(sentence) - 1)
if word in dic_ner:
if position != len(sentence) - 1:
seqner.write((dic_ner[word] + ' '))
else:
seqner.write((dic_ner[word]))
else: # 否则写入O
if position != len(sentence) - 1:
seqner.write('O' + ' ')
else:
seqner.write('O')
seqner.write('\n')
print('非空NER数据的句子占文件中所有句子的比例{:.1%}： '.format(j / len(sentence_list)), j, len(sentence_list))
return need_list
def ner_text_all(sentence_list, new_ner_path): # 不！！！去掉空NER行，生成新的数据集
j = 0 # j 用来统计全是空NER标签的句子数量，进而计算比例
need_list = []
with open(new_ner_path, 'w') as seqner:
for i, sentence in enumerate(sentence_list):
sentence_str = ' '.join(sentence)
doc = nlp(sentence_str)
dic_ner = {}
# 曾经不合适的字典
# for entity in doc.ents:
# for word in entity.text.split(' '):
# dic_ner[word] = entity.label_
# print(dic_ner)
for entity in doc.ents:
# print(entity.text, ':', entity.label_)
# print(len(entity.text.split(' ')))
# 一个单词的情况
if len(entity.text.split(' ')) == 1:
dic_ner[entity.text] = entity.label_
# 对于一个单词的情况
if len(entity.text.split(' ')) > 1:
for word in entity.text.split(' '):
if word == entity.text.split(' ')[0]:
dic_ner[word] = entity.label_
else:
dic_ner[word] = entity.label_
if doc.ents: # 如果这行数据有NER标签,则写入数据，否则这条数据忽略
# for entity in doc.ents:
# print(entity.text,entity.label_)
j = j + 1
need_list.append(i) # 有NER数据的时候就把这个数据的ID保存下来
for position, word in enumerate(sentence):
# print(position, len(sentence) - 1)
if word in dic_ner:
if position != len(sentence) - 1:
seqner.write((dic_ner[word] + ' '))
else:
seqner.write((dic_ner[word]))
else: # 否则写入O
if position != len(sentence) - 1:
seqner.write('O' + ' ')
else:
seqner.write('O')
seqner.write('\n')
else:
j = j + 1
need_list.append(i) # 有NER数据的时候就把这个数据的ID保存下来
for position, word in enumerate(sentence):
# print(position, len(sentence) - 1)
if word in dic_ner:
if position != len(sentence) - 1:
seqner.write((dic_ner[word] + ' '))
else:
seqner.write((dic_ner[word]))
else: # 否则写入O
if position != len(sentence) - 1:
seqner.write('O' + ' ')
else:
seqner.write('O')
seqner.write('\n')
print('非空NER数据的句子占文件中所有句子的比例{:.1%}： '.format(j / len(sentence_list)), j, len(sentence_list))
return need_list
def generat_new_data(need_list, old_list, new_path):
# print(old_list)
with open(new_path, 'w') as new:
for i, id in enumerate(need_list):
new.write(str((old_list[id])).strip() + '\n')
print("ok")
if __name__ == '__main__':
train_sentence_list = load_sentences4spacy(
os.path.join(base_dir, mid_middle, 'train/seq.in'))
valid_sentence_list = load_sentences4spacy(
os.path.join(base_dir, mid_middle, 'valid/seq.in'))
test_sentence_list = load_sentences4spacy(
os.path.join(base_dir, mid_middle, 'test/seq.in'))
if 'all' in new_middle:
# 生成新的数据seq.ner
print('all ', new_middle)
need_list_train = ner_text_all(train_sentence_list, new_train_ner_path)
need_list_valid = ner_text_all(valid_sentence_list, new_valid_ner_path)
need_list_test = ner_text_all(test_sentence_list, new_test_ner_path)
else:
print('part ', new_middle)
need_list_train = ner_text(train_sentence_list, new_train_ner_path)
need_list_valid = ner_text(valid_sentence_list, new_valid_ner_path)
need_list_test = ner_text(test_sentence_list, new_test_ner_path)
train_labels = load_labels(os.path.join(base_dir, middle, 'train/label'))
valid_labels = load_labels(os.path.join(base_dir, middle, 'valid/label'))
test_labels = load_labels(os.path.join(base_dir, middle, 'test/label'))
# # print(len(need_list_test), need_list_test)
generat_new_data(need_list_train, train_labels, new_train_label_path)
generat_new_data(need_list_valid, valid_labels, new_valid_label_path)
generat_new_data(need_list_test, test_labels, new_test_label_path)
train_in = load_labels(os.path.join(base_dir, middle, 'train/seq.in'))
valid_in = load_labels(os.path.join(base_dir, middle, 'valid/seq.in'))
test_in = load_labels(os.path.join(base_dir, middle, 'test/seq.in'))
generat_new_data(need_list_train, train_in, new_train_in_path)
generat_new_data(need_list_valid, valid_in, new_valid_in_path)
generat_new_data(need_list_test, test_in, new_test_in_path)
train_out = load_labels(os.path.join(base_dir, middle, 'train/seq.out'))
valid_out = load_labels(os.path.join(base_dir, middle, 'valid/seq.out'))
test_out = load_labels(os.path.join(base_dir, middle, 'test/seq.out'))
generat_new_data(need_list_train, train_out, new_train_out_path)
generat_new_data(need_list_valid, valid_out, new_valid_out_path)
generat_new_data(need_list_test, test_out, new_test_out_path)