forked from wcmac/sippycup
-
Notifications
You must be signed in to change notification settings - Fork 0
/
graph_kb.py
410 lines (365 loc) · 16.3 KB
/
graph_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
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
"""
This file defines two classes:
- GraphKB is a generic graph-structured knowledge base, or equivalently,
a set of relational pairs and triples, with indexing for fast lookups.
- GraphKBExecutor executes queries against a GraphKB. It defines a simple
query language, and responds to queries with (possibly empty) sets of
values from the GraphKB.
The primary use case for these classes within SippyCup is to provide an executor
backed by Geobase for use with the GeoQuery domain. However, these classes are
generic enough that they could be used for other applications. For example,
Freebase is also a graph-structured knowledge base.
See further documentation on the individual classes, in particular for a
detailed description of the query language defined by GraphKBExecutor.
See the demo() method at the bottom of this file for a concrete example.
"""
__author__ = "Bill MacCartney"
__copyright__ = "Copyright 2015, Bill MacCartney"
__credits__ = []
__license__ = "GNU General Public License, version 2.0"
__version__ = "0.9"
__maintainer__ = "Bill MacCartney"
__email__ = "See the author's website"
from collections import defaultdict, Iterable
from types import FunctionType
class GraphKB:
"""
Represents a knowledge base as set of tuples, each either:
- a pair, consisting of a unary relation and an element which belongs to it,
or
- a triple consisting of a binary relation and a pair of elements which
belong to it.
There are no restrictions on the types of the tuple elements, except that
they be indexable (hashable).
For example, a GraphKB could be constructed from this set of tuples:
('male', 'homer')
('female', 'marge')
('male', 'bart')
('female', 'lisa')
('female', 'maggie')
('adult', 'homer')
('adult', 'marge')
('child', 'bart')
('child', 'lisa')
('child', 'maggie')
('has_age', 'homer', 36)
('has_age', 'marge', 34)
('has_age', 'bart', 10)
('has_age', 'lisa', 8)
('has_age', 'maggie', 1)
('has_brother', 'lisa', 'bart')
('has_brother', 'maggie', 'bart')
('has_sister', 'bart', 'maggie')
('has_sister', 'bart', 'lisa')
('has_father', 'bart', 'homer')
('has_father', 'lisa', 'homer')
('has_father', 'maggie', 'homer')
('has_mother', 'bart', 'marge')
('has_mother', 'lisa', 'marge')
('has_mother', 'maggie', 'marge'),
"""
def __init__(self, tuples):
self.nodes = set()
self.unaries = defaultdict(set)
self.binaries_fwd = defaultdict(lambda : defaultdict(set)) # rel => src => {dst}
self.binaries_rev = defaultdict(lambda : defaultdict(set)) # rel => dst => {src}
for tuple in tuples:
if len(tuple) == 2:
self.add_unary(tuple)
elif len(tuple) == 3:
self.add_binary(tuple)
else:
assert False, 'Invalid tuple'
def add_unary(self, tuple):
self.nodes.add(tuple[1])
self.unaries[tuple[0]].add(tuple[1])
def add_binary(self, tuple):
self.nodes.add(tuple[1])
self.nodes.add(tuple[2])
self.binaries_fwd[tuple[0]][tuple[1]].add(tuple[2])
self.binaries_rev[tuple[0]][tuple[2]].add(tuple[1])
def list(self):
for rel in sorted(list(self.unaries.keys())):
for node in sorted(list(self.unaries[rel])):
print("(%s %s)" % (rel, node))
for rel in sorted(list(self.binaries_fwd.keys())):
for src in sorted(list(self.binaries_fwd[rel])):
for dst in sorted(list(self.binaries_fwd[rel][src])):
print("(%s %s %s)" % (rel, src, dst))
def executor(self):
return GraphKBExecutor(self)
class GraphKBExecutor:
"""
Executes formal queries against a GraphKB and returns their denotations.
Queries are represented by Python tuples, and can be nested.
Denotations are also represented by Python tuples, but are conceptually sets.
The query language is perhaps most easily explained by example:
query denotation
---------------------------------------- ------------------------
'bart' ('bart',)
'male' ('bart', 'homer')
('has_sister', 'lisa') ('bart', 'maggie')
('lisa', 'has_sister') ('maggie',)
('lisa', 'has_brother') ('bart',)
('.and', 'male', 'child') ('bart',)
('.or', 'male', 'adult') ('bart', 'homer', 'marge')
('.count', ('bart', 'has_sister')) (2,)
('has_age', ('.gt', 21)) ('homer', 'marge')
('has_age', ('.lt', 2)) ('maggie',)
('has_age', ('.eq', 10)) ('bart',)
('.argmax', 'has_age', 'female') ('marge',)
('.argmin', 'has_age', ('bart', 'has_sister')) ('maggie',)
A bit more formally: if
v is a value
U is a unary relation
B is a binary relation
Q is a relation having numeric values
X and Y are queries
[[X]] is the denotation of query X
then we can define the denotations of queries as follows:
query denotation
v the singleton set containing the value v
U the set of values belonging unary relation U
(B, X) the set of values which have relation B to any value in [[X]]
(X, B) the set of values to which any value in [[X]] has relation B
('.and', X, Y) the intersection of [[X]] and [[Y]]
('.or', X, Y) the union of [[X]] and [[Y]]
('.count', X) the cardinality of [[X]]
('.gt', X) the set of numbers greater than any number in [[X]]
('.lt', X) the set of numbers less than any number in [[X]]
('.eq', X) the set of numbers equal to the single number in [[X]]
('.argmax', Q, X) the subset of [[X]] having maximal values under relation Q
('.argmin', Q, X) the subset of [[X]] having minimal values under relation Q
"""
def __init__(self, graph_kb):
self.graph_kb = graph_kb
def execute(self, sem):
if isinstance(sem, tuple):
return self.execute_tuple(sem)
elif isinstance(sem, str) and sem.startswith('.'):
return self.execute_special((sem,))
elif sem in self.graph_kb.unaries:
return self.execute_unary(sem)
elif sem in self.graph_kb.binaries_fwd:
# It's a relation name, so return it as is.
return sem
else:
# It's some other value (string, integer, ...), so return it as a tuple.
return (sem,)
def execute_tuple(self, sem):
if len(sem) == 1 and sem[0] in self.graph_kb.unaries:
return self.execute_unary(sem[0])
elif len(sem) == 2 and sem[0] in self.graph_kb.binaries_fwd:
return self.execute_binary(sem[0], sem[1], rev=True)
elif len(sem) == 2 and sem[1] in self.graph_kb.binaries_fwd:
return self.execute_binary(sem[1], sem[0], rev=False)
elif sem[0].startswith('.'):
return self.execute_special(sem)
def execute_unary(self, rel):
return sorted_tuple(self.graph_kb.unaries[rel])
def execute_binary(self, rel, arg, rev=False):
arg = self.execute(arg)
index = self.graph_kb.binaries_rev if rev else self.graph_kb.binaries_fwd
if isinstance(arg, Iterable):
# arg is a tuple, e.g., the result of executing '5044'.
vals = [src for dst in arg for src in index[rel][dst]]
elif isinstance(arg, FunctionType):
# arg is a predicate, e.g., the result of executing ('.gt', 5044).
vals = [src for dst, srcs in list(index[rel].items()) if arg(dst) for src in srcs]
else:
raise Exception('Unsupported argument to join: %s' % str(arg))
return sorted_tuple(set(vals))
def execute_special(self, sem):
args = tuple([self.execute(elt) for elt in sem[1:]])
if sem[0] == '.and':
return self.execute_and(args)
elif sem[0] == '.or':
return self.execute_or(args)
elif sem[0] == '.not':
return self.execute_not(args)
elif sem[0] == '.any':
return self.execute_any(args)
elif sem[0] == '.count':
return self.execute_count(args)
elif sem[0] == '.gt':
return self.execute_gt(args)
elif sem[0] == '.lt':
return self.execute_lt(args)
elif sem[0] == '.eq':
return self.execute_eq(args)
elif sem[0] == '.max':
return self.execute_max(args, rev=False, arg=False)
elif sem[0] == '.min':
return self.execute_max(args, rev=True, arg=False)
elif sem[0] == '.argmax':
return self.execute_max(args, rev=False, arg=True)
elif sem[0] == '.argmin':
return self.execute_max(args, rev=True, arg=True)
else:
raise Exception('Unsupported operator: %s' % str(sem[0]))
def execute_and(self, args):
assert len(args) == 2
# Check to see if either element of args is a predicate,
# e.g., the result of executing ('.gt', 5044).
if isinstance(args[0], FunctionType):
args = (args[1], args[0])
if isinstance(args[1], FunctionType):
return sorted_tuple([elt for elt in args[0] if args[1](elt)])
else:
return sorted_tuple(set(args[0]).intersection(set(args[1])))
# TODO: Properly handle the case where one or both arguments are
# functions, like execute_and() does.
def execute_or(self, args):
assert len(args) == 2
return sorted_tuple(set(args[0]).union(set(args[1])))
def execute_not(self, args):
assert len(args) == 1
complement = [node for node in self.graph_kb.nodes if node not in args[0]]
return sorted_tuple(complement)
def execute_any(self, args):
assert len(args) == 0
return self.execute_not(((),))
def execute_count(self, args):
assert len(args) == 1
return (len(args[0]),)
def execute_gt(self, args):
assert len(args) == 1
assert isinstance(args[0], tuple), 'Not a tuple: %s' % str(args[0])
max_val = max(args[0]) if args[0] else float('-inf')
return lambda x: x > max_val
# TODO: consider ways of combining with execute_gt().
def execute_lt(self, args):
assert len(args) == 1
assert isinstance(args[0], tuple), 'Not a tuple: %s' % str(args[0])
min_val = min(args[0]) if args[0] else float('inf')
return lambda x: x < min_val
def execute_eq(self, args):
assert len(args) == 1
assert isinstance(args[0], tuple), 'Not a tuple: %s' % str(args[0])
assert len(args[0]) == 1
return lambda x: x == args[0][0]
def execute_max(self, args, rev=False, arg=False):
assert len(args) == 2
# TODO: Drop the assumption that the first argument is a relation from some entity
# to a number. What if it's the other way around?
index = self.graph_kb.binaries_fwd
assert args[0] in index, 'Not a relation name: %s' % str(args[0])
pairs = [(src, dst) for src in args[1] for dst in index[args[0]][src]]
vals = [val for e, val in pairs]
if rev:
ext_val = min(vals) if pairs else float('inf')
else:
ext_val = max(vals) if pairs else float('-inf')
if arg:
return sorted_tuple(set([e for e, val in pairs if val == ext_val]))
else:
return (ext_val,)
def sorted_tuple(elements):
return tuple(sorted(list(elements), key=str))
# demo =========================================================================
def demo():
from example import Example
tuples = [
('male', 'homer'),
('female', 'marge'),
('male', 'bart'),
('female', 'lisa'),
('female', 'maggie'),
('adult', 'homer'),
('adult', 'marge'),
('child', 'bart'),
('child', 'lisa'),
('child', 'maggie'),
('has_age', 'homer', 36),
('has_age', 'marge', 34),
('has_age', 'bart', 10),
('has_age', 'lisa', 8),
('has_age', 'maggie', 1),
('has_brother', 'lisa', 'bart'),
('has_brother', 'maggie', 'bart'),
('has_sister', 'bart', 'maggie'),
('has_sister', 'bart', 'lisa'),
('has_father', 'bart', 'homer'),
('has_father', 'lisa', 'homer'),
('has_father', 'maggie', 'homer'),
('has_mother', 'bart', 'marge'),
('has_mother', 'lisa', 'marge'),
('has_mother', 'maggie', 'marge'),
]
graph_kb = GraphKB(tuples)
graph_kb.list()
executor = graph_kb.executor()
examples = [
Example(input='bart',
semantics='bart',
denotation=('bart',)),
Example(input='males',
semantics='male',
denotation=('bart', 'homer')),
Example(input='things that have bart as brother',
semantics=('has_brother', 'bart'),
denotation=('lisa', 'maggie')),
Example(input='brothers of bart',
semantics=('bart', 'has_brother'),
denotation=()),
Example(input='sisters of bart',
semantics=('bart', 'has_sister'),
denotation=('lisa', 'maggie')),
Example(input='male children',
semantics=('.and', 'male', 'child'),
denotation=('bart',)),
Example(input='males or adults',
semantics=('.or', 'male', 'adult'),
denotation=('bart', 'homer', 'marge')),
Example(input='not a child',
semantics=('.not', 'child'),
denotation=(1, 10, 34, 36, 8, 'homer', 'marge')),
Example(input='anything',
semantics=('.any',),
denotation=(1, 10, 34, 36, 8, 'bart', 'homer', 'lisa', 'maggie', 'marge')),
Example(input='sisters',
semantics=('.any', 'has_sister'),
denotation=('lisa', 'maggie')),
Example(input='children who are not sisters',
semantics=('.and', 'child', ('.not', ('.any', 'has_sister'))),
denotation=('bart',)),
Example(input='number of sisters of bart',
semantics=('.count', ('bart', 'has_sister')),
denotation=(2,)),
Example(input='things that have age greater than 21',
semantics=('has_age', ('.gt', 21)),
denotation=('homer', 'marge')),
Example(input='things that have age less than 2',
semantics=('has_age', ('.lt', 2)),
denotation=('maggie',)),
Example(input='things that have age equal to 10',
semantics=('has_age', ('.eq', 10)),
denotation=('bart',)),
Example(input='age of oldest female',
semantics=('.max', 'has_age', 'female'),
denotation=(34,)),
Example(input='age of youngest sister of bart',
semantics=('.min', 'has_age', ('bart', 'has_sister')),
denotation=(1,)),
Example(input='age of oldest thing',
semantics=('.max', 'has_age', '.any'),
denotation=(36,)),
Example(input='oldest female',
semantics=('.argmax', 'has_age', 'female'),
denotation=('marge',)),
Example(input='youngest sister of bart',
semantics=('.argmin', 'has_age', ('bart', 'has_sister')),
denotation=('maggie',)),
Example(input='oldest thing',
semantics=('.argmax', 'has_age', '.any'),
denotation=('homer',)),
]
for example in examples:
deno = executor.execute(example.semantics)
assert deno == example.denotation, example.input + ': ' + str(deno)
print()
print('%-16s %s' % ('input', example.input))
print('%-16s %s' % ('semantics', example.semantics))
print('%-16s %s' % ('denotation', deno))
if __name__ == '__main__':
demo()