-
Notifications
You must be signed in to change notification settings - Fork 4
/
Copy pathdata_generate.py
158 lines (121 loc) · 7.55 KB
/
data_generate.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
import os
import argparse
import pickle
import numpy as np
from utils.data_utils import load_raw_data, read_triplets2id, count_entity_relation_set, write_triplets
def load_grail_data(dataset, version, entity2id, relation2id):
"""读取grail的数据
Args:
dataset (str): 读取哪个数据集
version (int): 读取v几的数据
entity2id (dict): entity到id的映射字典
relation2id (dict): relation到id的映射字典
Returns:
original_triplets (list): original的所有triplets，二维列表
original_entity_set (set): original中的所有entity
original_relation_set (set): original中的所有relation
emerging_triplets (list): emerging的所有triplets，二维列表
emerging_entity_set (set): emerging中的所有entity
emerging_relation_set (set): emerging中的所有relation
"""
original_data_path = f'./data/grail/{dataset}_v{version}'
emerging_data_path = f'./data/grail/{dataset}_v{version}_ind'
original_train_triplets, entity2id, relation2id = read_triplets2id(os.path.join(original_data_path, 'train.txt'), entity2id, relation2id)
original_valid_triplets, entity2id, relation2id = read_triplets2id(os.path.join(original_data_path, 'valid.txt'), entity2id, relation2id)
original_test_triplets, entity2id, relation2id = read_triplets2id(os.path.join(original_data_path, 'test.txt'), entity2id, relation2id)
original_triplets = original_train_triplets + original_valid_triplets + original_test_triplets
original_entity_set , original_relation_set = count_entity_relation_set(original_triplets)
emerging_train_triplets, entity2id, relation2id = read_triplets2id(os.path.join(emerging_data_path, 'train.txt'), entity2id, relation2id)
emerging_valid_triplets, entity2id, relation2id = read_triplets2id(os.path.join(emerging_data_path, 'valid.txt'), entity2id, relation2id)
emerging_test_triplets, entity2id, relation2id = read_triplets2id(os.path.join(emerging_data_path, 'test.txt'), entity2id, relation2id)
emerging_triplets = emerging_train_triplets + emerging_valid_triplets + emerging_test_triplets
emerging_entity_set, emerging_relation_set = count_entity_relation_set(emerging_triplets)
return original_train_triplets, original_valid_triplets, original_test_triplets, original_entity_set, original_relation_set, \
emerging_train_triplets, emerging_valid_triplets, emerging_test_triplets, emerging_entity_set, emerging_relation_set, \
entity2id, relation2id
def screen_u2s_triplets(raw_triplets, original_entity_set, emerging_entity_set):
u2s_triplets = []
for triplet in raw_triplets:
h, r, t = triplet
if h in original_entity_set and t in emerging_entity_set:
u2s_triplets.append(triplet)
elif t in original_entity_set and h in emerging_entity_set:
u2s_triplets.append(triplet)
return u2s_triplets
def main(params):
# 读取raw数据
entity2id, relation2id, train_triplets, valid_triplets, test_triplets = load_raw_data('./data/{}'.format(params.dataset))
all_triplets = train_triplets + valid_triplets + test_triplets
entity_set = set(range(len(entity2id)))
relation_set = set(range(len(relation2id)))
version = [1, 2, 3]
for v in version:
# 读取grail中的数据
# 注意，这里处理数据时发现了一个问题，grail提供的entity2id和relation2id，在nell和wn上均出现了不全的情况，因此这里重新统计并做了补全
original_train_triplets, original_valid_triplets, original_test_triplets, original_entity_set, original_relation_set, \
emerging_train_triplets, emerging_valid_triplets, emerging_test_triplets, emerging_entity_set, emerging_relation_set, \
entity2id, relation2id = load_grail_data(params.dataset, v, entity2id, relation2id)
# 也更新一下id2
id2entity = {v: k for k, v in entity2id.items()}
id2relation = {v: k for k, v in relation2id.items()}
print('For dataset {}_v{}, there are {} original train triplets, {} emerging train triplets.'.format(params.dataset, v, len(original_train_triplets), len(emerging_train_triplets)))
# 筛选所有u2s数据
u2s_triplets = screen_u2s_triplets(all_triplets, original_entity_set, emerging_entity_set)
# 确保所有u2s数据中的relation是已知的
u2s_triplets = np.array([triplet for triplet in u2s_triplets if triplet[1] in original_relation_set])
print("There are {} triplets as the u2s triplets".format(len(u2s_triplets)))
train_triplets = np.array(original_train_triplets + emerging_train_triplets)
np.random.shuffle(train_triplets)
print("There are {} triplets as the all train triplets.".format(len(train_triplets)))
# v1 对应构建 EQ, enclosing links 和 bridging links 比例为1:1
# v2 对应构建 MB, enclosing links 和 bridging links 比例为1:2
# v3 对应构建 ME, enclosing links 和 bridging links 比例为2:1
if v == 1:
num_bri = len(emerging_test_triplets)
l = "EQ"
elif v == 2:
num_bri = int(len(emerging_test_triplets) * 2)
l = "MB"
else:
num_bri = int(len(emerging_test_triplets) / 2)
l = "ME"
if num_bri > len(u2s_triplets):
num_bri = len(u2s_triplets)
# 采样
bri_idx = np.random.choice(np.array(range(len(u2s_triplets))), num_bri, False)
bri_triplets = u2s_triplets[bri_idx]
np.random.shuffle(bri_triplets)
print("Sample {} triplets as bridge triplets.\n".format(len(valid_triplets)))
# 写原始数据
data_path = './data/{}_{}'.format(params.dataset, l)
if not os.path.exists(data_path):
os.makedirs(data_path)
write_triplets(os.path.join(data_path, 'train.txt'), original_train_triplets, id2entity, id2relation)
write_triplets(os.path.join(data_path, 'valid.txt'), original_valid_triplets, id2entity, id2relation)
# 写enc数据
data_path_enc = './data/{}_{}_enc'.format(params.dataset, l)
if not os.path.exists(data_path_enc):
os.makedirs(data_path_enc)
write_triplets(os.path.join(data_path_enc, 'train.txt'), emerging_train_triplets, id2entity, id2relation)
write_triplets(os.path.join(data_path_enc, 'test.txt'), emerging_test_triplets, id2entity, id2relation)
# 写bri数据
data_path_bri = './data/{}_{}_bri'.format(params.dataset, l)
if not os.path.exists(data_path_bri):
os.makedirs(data_path_bri)
write_triplets(os.path.join(data_path_bri, 'train.txt'), train_triplets, id2entity, id2relation)
write_triplets(os.path.join(data_path_bri, 'test.txt'), bri_triplets, id2entity, id2relation)
# 写mix数据
data_path_mix = './data/{}_{}_mix'.format(params.dataset, l)
mix_triplets = np.concatenate((bri_triplets, emerging_test_triplets))
np.random.shuffle(mix_triplets)
if not os.path.exists(data_path_mix):
os.makedirs(data_path_mix)
write_triplets(os.path.join(data_path_mix, 'train.txt'), train_triplets, id2entity, id2relation)
write_triplets(os.path.join(data_path_mix, 'test.txt'), mix_triplets, id2entity, id2relation)
if __name__ == '__main__':
parser = argparse.ArgumentParser(description='data_generate')
# 数据集设置
parser.add_argument('--dataset', '-d', type=str,
help='用哪个数据集')
params = parser.parse_args()
main(params)