/
split_data.py
306 lines (225 loc) · 13.3 KB
/
split_data.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
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
import random
import spacy
import pandas as pd
import os
nlp_en = spacy.load('en_core_web_md')
from read_write_file import *
def get_text_by_length(content, instance_text, max_length = 256, margin = 0):
if (instance_text == ''):
con_list = content.split()
con_list = con_list[0:max_length]
content = ' '.join(w for w in con_list)
if (content[-1] != '.'): content += '.'
return content
con_list = content.split()
ins_list = instance_text.split()
con_length = max_length - margin - len(ins_list)
if (con_length < 0):
con_list = content.split()
con_list = con_list[0:max_length]
content = ' '.join(w for w in con_list)
if (content[-1] != '.'): content += '.'
return content
con_list = con_list[0:con_length]
con_text = ' '.join(w for w in con_list)
if (con_text[-1] != '.'): con_text += '.'
ins_text = ' '.join(w for w in ins_list)
return con_text + ' ' + ins_text
def split_phrase1_dataset(input_file = 'dataset/collected_data.json', format_json = False, \
des_type = 'para_wd', max_length = 256, margin = 16):
dataset = load_list_from_json_file(input_file, format_json = format_json)
data_list = []
for item in dataset:
# filter bad examples
if ('wiki' in item['label'].lower()): continue # remove Wikimedia items
if ('wiki' in item['description'].lower()): continue # remove Wikimedia items
if (item['first_sentence'].strip() == ''): continue # remove empty first sentences
if (len(item['first_sentence'].split()) < 10): continue # remove short first sentences
if (len(item['instances']) == 0): continue # remove empty instances
description = item['description']
first_sentence = item['first_sentence']
first_paragraph = item['first_paragraph']
label = item['label']
if (len(label) == 0): continue
if (len(label) == 1): label = label.upper()
else: label = label[0].upper() + label[1:]
instances = item['instances']
instances = [i for i in instances if i[1].lower() != item['label'].lower()] # remove the label in instances
subclasses = item['subclasses']
subclasses = [i for i in subclasses if i[1].lower() != item['label'].lower()] # remove the label in subclasses
instance_text = ''
if (len(instances) == 1): instance_text = label + ' is a ' + instances[0][1] + '.'
elif (len(instances) == 2): instance_text = label + ' is a ' + instances[0][1] + ' and ' + instances[1][1] + '.'
elif (len(instances) > 1):
instance_text = ', '.join(str(i[1]) for i in instances[0:len(instances)-1] if str(i).strip() != '')
instance_text = label + ' is a ' + instance_text
instance_text += ', and ' + instances[-1][1] + '.'
source = ''
if (des_type == 'sen_wd'): # first sentence + wikidata instances
#source = first_sentence
#if (instance_text != ''): source += ' ' + instance_text
source = get_text_by_length(first_sentence, instance_text, max_length = max_length)
elif(des_type == 'para_wd'): # first sentence + wikidata instances
#source = first_paragraph
#if (instance_text != ''): source += ' ' + instance_text
source = get_text_by_length(first_paragraph, instance_text, max_length = max_length)
elif(des_type == 'para'): # first sentence only
#source = first_paragraph
con_list = first_paragraph.split()
con_list = con_list[0:max_length - margin]
source = ' '.join(w for w in con_list)
else:
#source = first_sentence
con_list = first_sentence.split()
con_list = con_list[0:max_length - margin]
source = ' '.join(w for w in con_list)
source = ' '.join(w for w in [w for w in source.split() if w.strip() != ''])
if (source[-1] != '.'): source += '.' # more elegant
# create candidate list
candidate_list = []
if (len(instances) != 0): candidate_list += [i[1] for i in instances]
if (len(subclasses) != 0): candidate_list += [i[1] for i in subclasses]
candidate_list = list(set(candidate_list))
if (source != ''):
data_list.append({'wikidata_id': item['wikidata_id'], 'label': item['label'], 'source': source, 'target': description, 'baseline_candidates':candidate_list})
# split into training, validation, and test sets with ratio 8:1:1
training_list = data_list[:(len(data_list)//10)*8]
validation_list = data_list[(len(data_list)//10)*8:]
test_list = validation_list[:(len(validation_list)//10)*5]
validation_list = validation_list[(len(validation_list)//10)*5:]
write_list_to_jsonl_file('dataset/phrase1/random/training_' + des_type + '_' + str(max_length) + '.json', \
training_list, file_access = 'w')
write_list_to_jsonl_file('dataset/phrase1/random/validation_' + des_type + '_' + str(max_length) + '.json', \
validation_list, file_access = 'w')
write_list_to_jsonl_file('dataset/phrase1/random/test_' + des_type + '_' + str(max_length) + '.json', \
test_list, file_access = 'w')
def split_phrase1_dataset2(training_instances, input_file = 'dataset/collected_data.json', format_json = False, \
des_type = 'para_wd', max_length = 256, margin = 0):
dataset = load_list_from_json_file(input_file, format_json = format_json)
data_list = []
print('training_instances: ', training_instances)
training_list, validation_list, test_list = [], [], []
for item in dataset:
# filter bad examples
if ('wiki' in item['label'].lower()): continue # remove Wikimedia items
if ('wiki' in item['description'].lower()): continue # remove Wikimedia items
if (item['first_sentence'].strip() == ''): continue # remove empty first sentences
if (len(item['first_sentence'].split()) < 10): continue # remove short first sentences
if (len(item['instances']) == 0): continue # remove empty instances
description = item['description']
first_sentence = item['first_sentence']
first_paragraph = item['first_paragraph']
label = item['label']
if (len(label) == 0): continue
if (len(label) == 1): label = label.upper()
else: label = label[0].upper() + label[1:]
instances = item['instances']
instances = [i for i in instances if i[1].lower() != item['label'].lower()] # remove the label in instances
#if ('Wikimedia list article' in instances): continue # pass the list articles
subclasses = item['subclasses']
subclasses = [i for i in subclasses if i[1].lower() != item['label'].lower()] # remove the label in subclasses
instance_text = ''
if (len(instances) == 1): instance_text = label + ' is a ' + instances[0][1] + '.'
elif (len(instances) == 2): instance_text = label + ' is a ' + instances[0][1] + ' and ' + instances[1][1] + '.'
elif (len(instances) > 1):
instance_text = ', '.join(str(i[1]) for i in instances[0:len(instances)-1] if str(i).strip() != '')
instance_text = label + ' is a ' + instance_text
instance_text += ', and ' + instances[-1][1] + '.'
source = ''
if (des_type == 'sen_wd'): # first sentence + wikidata instances
#source = first_sentence
#if (instance_text != ''): source += ' ' + instance_text
source = get_text_by_length(first_sentence, instance_text, max_length = max_length)
elif(des_type == 'para_wd'): # first sentence + wikidata instances
#source = first_paragraph
#if (instance_text != ''): source += ' ' + instance_text
source = get_text_by_length(first_paragraph, instance_text, max_length = max_length)
elif(des_type == 'para'): # first sentence only
#source = first_paragraph
con_list = first_paragraph.split()
con_list = con_list[0:max_length - margin]
source = ' '.join(w for w in con_list)
else:
#source = first_sentence
con_list = first_sentence.split()
con_list = con_list[0:max_length - margin]
source = ' '.join(w for w in con_list)
source = ' '.join(w for w in [w for w in source.split() if w.strip() != ''])
if (source[-1] != '.'): source += '.' # more elegant
# create candidate list
candidate_list = []
if (len(instances) != 0): candidate_list += [i[1] for i in instances]
#if (len(subclasses) != 0): candidate_list += [i[1] for i in subclasses]
candidate_list = list(set(candidate_list))
if (source != ''):
check = check_common_item(instances, training_instances)
if (check == True):
training_list.append({'wikidata_id': item['wikidata_id'], 'label': item['label'], \
'source': source, 'target': description, 'baseline_candidates':candidate_list})
else:
validation_list.append({'wikidata_id': item['wikidata_id'], 'label': item['label'], \
'source': source, 'target': description, 'baseline_candidates':candidate_list})
# split 5:5
test_list = validation_list[(len(validation_list)//10)*5:]
validation_list = validation_list[:(len(validation_list)//10)*5]
print('training, validation, test: ', len(training_list), len(validation_list), len(test_list))
write_list_to_jsonl_file('dataset/phrase1/diff/training_' + des_type + '_' + str(max_length) + '.json', \
training_list, file_access = 'w')
write_list_to_jsonl_file('dataset/phrase1/diff/validation_' + des_type + '_' + str(max_length) + '.json', \
validation_list, file_access = 'w')
write_list_to_jsonl_file('dataset/phrase1/diff/test_' + des_type + '_' + str(max_length) + '.json', \
test_list, file_access = 'w')
def check_common_item(list1, list2):
for i in list1:
if (i[1] in list2):
return True
return False
def split_phrase1_setup(input_file = 'dataset/collected_data.json'):
max_list = [32, 64, 128, 256, 512, 1024]
margin_list = [0, 0, 0, 0, 0, 0]
instance_dict, dataset = instance_distribution(input_file)
training_instances = get_training_instance(instance_dict, dataset)
# create folders
if not os.path.exists('dataset/phrase1/diff'): os.makedirs('dataset/phrase1/diff')
if not os.path.exists('dataset/phrase1/random'): os.makedirs('dataset/phrase1/random')
for x, y in zip(max_list, margin_list):
#split_dataset(des_type = 'para_wd', max_length = l)
#split_dataset(des_type = 'sen_wd', max_length = l)
#split_dataset(des_type = 'sen', max_length = l)
split_phrase1_dataset2(training_instances, des_type = 'para', max_length = x, margin = y)
split_phrase1_dataset(des_type = 'para', max_length = x, margin = y)
def split_phrase2_dataset(input_file = 'dataset/collected_sum.json', format_json = True, max_length = 128):
dataset = load_list_from_jsonl_file(input_file)
training_list = dataset[:(len(dataset)//10)*8]
validation_list = dataset[(len(dataset)//10)*8:]
test_list = validation_list[(len(validation_list)//10)*5:]
validation_list = validation_list[:(len(validation_list)//10)*5]
write_list_to_jsonl_file('dataset/phrase2/training_sum.json', training_list, file_access = 'w')
write_list_to_jsonl_file('dataset/phrase2/validation_sum.json', validation_list, file_access = 'w')
write_list_to_jsonl_file('dataset/phrase2/test_sum.json', test_list, file_access = 'w')
def instance_distribution(input_file = 'dataset/collected_data.json'):
dataset = load_list_from_json_file(input_file, format_json = False)
print(len(dataset))
instance_dict = {}
for item in dataset:
instances = item['instances']
for i in instances:
if (i[1] not in instance_dict):
instance_dict[i[1]] = 1
else:
instance_dict[i[1]] += 1
instance_dict = dict(sorted(instance_dict.items(), key = lambda x:x[1], reverse = True))
write_list_to_json_file('dataset/instance_distribution.json', instance_dict, file_access = 'w')
return instance_dict, dataset
def get_training_instance(instance_dict, dataset, training_rate = 0.8):
training_len = int(len(dataset)*training_rate)
training_instances = []
count = 0
for item in instance_dict.items():
if (count >= training_len): break
count += item[1]
training_instances.append(item[0])
return training_instances
#.........................................
if __name__ == "__main__":
split_phrase1_setup()