forked from shijx12/KQAPro_Baselines
-
Notifications
You must be signed in to change notification settings - Fork 0
/
load_kb.py
324 lines (300 loc) · 11.8 KB
/
load_kb.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
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
import json
from collections import Counter
from utils.misc import init_vocab
from datetime import date
from queue import Queue
from utils.value_class import ValueClass
"""
knowledge json format:
'concepts':
{
'id':
{
'name': '',
'instanceOf': ['<concept_id>'],
}
},
'entities': # exclude concepts
{
'id':
{
'name': '<entity_name>',
'instanceOf': ['<concept_id>'],
'attributes':
[
{
'key': '<key>',
'value':
{
'type': 'string'/'quantity'/'date'/'year'
'value': # float or int for quantity, int for year, 'yyyy/mm/dd' for date
'unit': # for quantity
},
'qualifiers':
{
'<qk>':
[
<qv>, # each qv is a dictionary like value, including keys type,value,unit
]
}
}
]
'relations':
[
{
'predicate': '<predicate>',
'object': '<object_id>', # NOTE: it may be a concept id
'direction': 'forward' or 'backward',
'qualifiers':
{
'<qk>':
[
<qv>, # each qv is a dictionary like value
]
}
}
]
}
}
"""
def get_kb_vocab(kb_json, min_cnt=1):
counter = Counter()
kb = json.load(open(kb_json))
for i in kb['concepts']:
counter.update([i, kb['concepts'][i]['name']])
for i in kb['entities']:
counter.update([i, kb['entities'][i]['name']])
for attr_dict in kb['entities'][i]['attributes']:
counter.update([attr_dict['key']])
values = [attr_dict['value']]
for qk, qvs in attr_dict['qualifiers'].items():
counter.update([qk])
values += qvs
for value in values:
u = value.get('unit', '')
if u:
counter.update([u])
counter.update([str(value['value'])])
for rel_dict in kb['entities'][i]['relations']:
counter.update([rel_dict['predicate'], rel_dict['direction']])
values = []
for qk, qvs in rel_dict['qualifiers'].items():
counter.update([qk])
values += qvs
for value in values:
u = value.get('unit', '')
if u:
counter.update([u])
counter.update([str(value['value'])])
vocab = init_vocab()
for v, c in counter.items():
if v and c >= min_cnt and v not in vocab:
vocab[v] = len(vocab)
return kb, vocab
def load_as_graph(kb_json, max_desc=200, min_cnt=1):
kb, vocab = get_kb_vocab(kb_json, min_cnt)
id2idx = {}
pred2idx = {}
node_descs = []
triples = []
for i, info in kb['concepts'].items():
id2idx[i] = len(id2idx)
desc = [info['name']]
node_descs.append(desc)
for i, info in kb['entities'].items():
id2idx[i] = len(id2idx)
desc = [info['name']]
for attr_info in info['attributes']:
desc.append(attr_info['key'])
desc.append(str(attr_info['value']['value']))
u = attr_info['value'].get('unit', '')
if u:
desc.append(u)
node_descs.append(desc)
for rel_info in info['relations']:
obj_id = rel_info['object']
if obj_id not in id2idx:
continue
pred = rel_info['predicate']
if pred not in pred2idx:
pred2idx[pred] = len(pred2idx)
pred_idx = pred2idx[pred]
sub_idx = id2idx[i]
obj_idx = id2idx[obj_id]
if rel_info['direction'] == 'forward':
triples.append((sub_idx, pred_idx, obj_idx))
else:
triples.append((obj_idx, pred_idx, sub_idx))
# encode and pad desc
for i, desc in enumerate(node_descs):
desc = [vocab.get(w, vocab['<UNK>']) for w in desc]
while len(desc) < max_desc:
desc.append(vocab['<PAD>'])
node_descs[i] = desc[:max_desc]
return vocab, node_descs, triples, id2idx, pred2idx
def load_as_key_value(kb_json, min_cnt=1):
"""
For KVMemNN
Load each triple (s, r, o) as kv pairs (s+r, o) and (o+r_, s)
"""
keys = ['<PAD>'] # use <PAD> as the first key
values = ['<PAD>']
def add_sro(s, r, o):
keys.append('{} {}'.format(s, r))
values.append(o)
keys.append('{} {}_'.format(o, r))
values.append(s)
kb = json.load(open(kb_json))
for i in kb['concepts']:
for j in kb['concepts'][i]['instanceOf']:
s = kb['concepts'][i]['name']
o = kb['concepts'][j]['name']
add_sro(s, 'instanceOf', o)
for i in kb['entities']:
for j in kb['entities'][i]['instanceOf']:
s = kb['entities'][i]['name']
o = kb['concepts'][j]['name']
add_sro(s, 'instanceOf', o)
name = kb['entities'][i]['name']
for attr_dict in kb['entities'][i]['attributes']:
o = '{} {}'.format(attr_dict['value']['value'], attr_dict['value'].get('unit', ''))
add_sro(name, attr_dict['key'], o)
s = '{} {} {}'.format(name, attr_dict['key'], o)
for qk, qvs in attr_dict['qualifiers'].items():
for qv in qvs:
o = '{} {}'.format(qv['value'], qv.get('unit', ''))
add_sro(s, qk, o)
for rel_dict in kb['entities'][i]['relations']:
if rel_dict['direction'] == 'backward': # we add reverse relation in add_sro
continue
o = kb['entities'].get(rel_dict['object'], kb['concepts'].get(rel_dict['object'], None))
if o is None: # wtf, why are some objects not in kb?
continue
o = o['name']
add_sro(name, rel_dict['predicate'], o)
s = '{} {} {}'.format(name, rel_dict['predicate'], o)
for qk, qvs in rel_dict['qualifiers'].items():
for qv in qvs:
o = '{} {}'.format(qv['value'], qv.get('unit', ''))
add_sro(s, qk, o)
print('length of kv pairs: {}'.format(len(keys)))
counter = Counter()
for i in range(len(keys)):
keys[i] = keys[i].lower().split()
values[i] = values[i].lower().split()
counter.update(keys[i])
counter.update(values[i])
vocab = init_vocab()
for v, c in counter.items():
if v and c >= min_cnt and v not in vocab:
vocab[v] = len(vocab)
return vocab, keys, values
class DataForSPARQL(object):
def __init__(self, kb_path):
kb = json.load(open(kb_path))
self.concepts = kb['concepts']
self.entities = kb['entities']
# replace adjacent space and tab in name, which may cause errors when building sparql query
for con_id, con_info in self.concepts.items():
con_info['name'] = ' '.join(con_info['name'].split())
for ent_id, ent_info in self.entities.items():
ent_info['name'] = ' '.join(ent_info['name'].split())
# get all attribute keys and predicates
self.attribute_keys = set()
self.predicates = set()
self.key_type = {}
for ent_id, ent_info in self.entities.items():
for attr_info in ent_info['attributes']:
self.attribute_keys.add(attr_info['key'])
self.key_type[attr_info['key']] = attr_info['value']['type']
for qk in attr_info['qualifiers']:
self.attribute_keys.add(qk)
for qv in attr_info['qualifiers'][qk]:
self.key_type[qk] = qv['type']
for ent_id, ent_info in self.entities.items():
for rel_info in ent_info['relations']:
self.predicates.add(rel_info['predicate'])
for qk in rel_info['qualifiers']:
self.attribute_keys.add(qk)
for qv in rel_info['qualifiers'][qk]:
self.key_type[qk] = qv['type']
self.attribute_keys = list(self.attribute_keys)
self.predicates = list(self.predicates)
# Note: key_type is one of string/quantity/date, but date means the key may have values of type year
self.key_type = { k:v if v!='year' else 'date' for k,v in self.key_type.items() }
# parse values into ValueClass object
for ent_id, ent_info in self.entities.items():
for attr_info in ent_info['attributes']:
attr_info['value'] = self._parse_value(attr_info['value'])
for qk, qvs in attr_info['qualifiers'].items():
attr_info['qualifiers'][qk] = [self._parse_value(qv) for qv in qvs]
for ent_id, ent_info in self.entities.items():
for rel_info in ent_info['relations']:
for qk, qvs in rel_info['qualifiers'].items():
rel_info['qualifiers'][qk] = [self._parse_value(qv) for qv in qvs]
def _parse_value(self, value):
if value['type'] == 'date':
x = value['value']
p1, p2 = x.find('/'), x.rfind('/')
y, m, d = int(x[:p1]), int(x[p1+1:p2]), int(x[p2+1:])
result = ValueClass('date', date(y, m, d))
elif value['type'] == 'year':
result = ValueClass('year', value['value'])
elif value['type'] == 'string':
result = ValueClass('string', value['value'])
elif value['type'] == 'quantity':
result = ValueClass('quantity', value['value'], value['unit'])
else:
raise Exception('unsupport value type')
return result
def get_direct_concepts(self, ent_id):
"""
return the direct concept id of given entity/concept
"""
if ent_id in self.entities:
return self.entities[ent_id]['instanceOf']
elif ent_id in self.concepts:
return self.concepts[ent_id]['instanceOf']
else:
raise Exception('unknown id')
def get_all_concepts(self, ent_id):
"""
return a concept id list
"""
ancestors = []
q = Queue()
for c in self.get_direct_concepts(ent_id):
q.put(c)
while not q.empty():
con_id = q.get()
ancestors.append(con_id)
for c in self.concepts[con_id]['instanceOf']:
q.put(c)
return ancestors
def get_name(self, ent_id):
if ent_id in self.entities:
return self.entities[ent_id]['name']
elif ent_id in self.concepts:
return self.concepts[ent_id]['name']
else:
return None
def is_concept(self, ent_id):
return ent_id in self.concepts
def get_attribute_facts(self, ent_id, key=None, unit=None):
if key:
facts = []
for attr_info in self.entities[ent_id]['attributes']:
if attr_info['key'] == key:
if unit:
if attr_info['value'].unit == unit:
facts.append(attr_info)
else:
facts.append(attr_info)
else:
facts = self.entities[ent_id]['attributes']
facts = [(f['key'], f['value'], f['qualifiers']) for f in facts]
return facts
def get_relation_facts(self, ent_id):
facts = self.entities[ent_id]['relations']
facts = [(f['predicate'], f['object'], f['direction'], f['qualifiers']) for f in facts]
return facts