-
Notifications
You must be signed in to change notification settings - Fork 1
/
parser_util.py
1812 lines (1427 loc) · 70.5 KB
/
parser_util.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
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#coding:utf-8
from feature import Feature
import util, config, os, json
from connective_dict import Connectives_dict
from example import Example
import model_trainer.mallet_util as mallet_util
from model_trainer.NT_arg_extractor.constituent import Constituent
from syntax_tree import Syntax_tree
from connective import Connective
from clause import Arg_Clauses
import copy
import sys
from model_trainer.cnn_implicit_classifier import cnn_config
def get_all_connectives(documents):
conns_list = [] #[(DocID, sent_index, conn_indices), ()..]
for DocID in documents:
doc = documents[DocID]
list = _get_doc_conns(doc) #[(sent_index, conn_indices), ()..]
for sent_index, conn_indices in list:
conns_list.append((DocID, sent_index, conn_indices))
conns_list = sorted(conns_list, key=lambda x: x[0])
return conns_list
def _get_doc_conns(document):
list = [] #[(sent_index, conn_indices), ()..]
for sent_index, sentence in enumerate(document["sentences"]):
sent_words_list = [word[0] for word in sentence["words"]]
for conn_indices in _check_connectives(sent_words_list): #[[2, 3], [0]]
list.append((sent_index, conn_indices))
return list
# identify connectives in sentence (sent_tokens)
# return indices: [[2, 3], [0]]
def _check_connectives(sent_tokens):
sent_tokens = [word.lower() for word in sent_tokens ]
indices = []
tagged = set([])
sortedConn = Connectives_dict().sorted_conns_list
for conn in sortedConn:
if '..' in conn:
c1, c2 = conn.split('..')
c1_indice = util.getSpanIndecesInSent(c1.split(), sent_tokens)#[[7]]
c2_indice = util.getSpanIndecesInSent(c2.split(), sent_tokens)#[[10]]
if c1_indice!= [] and c2_indice != []:
if c1_indice[0][0] < c2_indice[0][0]:
temp = set([t for t in (c1_indice[0]+c2_indice[0]) ])
if tagged & temp == set([]):
indices.append(c1_indice[0]+c2_indice[0])# [[7], [10]]
tagged = tagged.union(temp)
else:
c_indice = util.getSpanIndecesInSent(conn.split(), sent_tokens)#[[2,6],[1,3],...]
if c_indice !=[]:
tt = []
for item in c_indice:
if set(item) & tagged == set([]):
tt.append(item)
c_indice = tt
if c_indice != []:
indices.extend([item for item in c_indice])#[([2,6], 'for instance'), ....]
tagged = tagged.union(set([r for t in c_indice for r in t]))
return indices
#[(DocID, sent_index, conn_indices), ()..]
def conn_clf_print_feature(parse_dict, conns_list, feature_function, to_file):
print("\tExtract features: [..]", end=' ')
example_list = []
for DocID, sent_index, conn_indices in conns_list:
feature = feature_function(parse_dict, DocID, sent_index, conn_indices)
example = Example("", feature)
example_list.append(example)
# write example_list to file
util.write_example_list_to_file(example_list, to_file)
print("\r\tExtract features: [OK]")
def conn_clf_read_model_output(conn_clf_model_output, conns_list):
# ['yes', 'no'...]
pred_list = mallet_util.get_mallet_predicted_list(conn_clf_model_output)
disc_conns = []
for pred, conn in zip(pred_list, conns_list):
if pred == "1":
disc_conns.append(conn)
return disc_conns
def arg_position_print_feature(parse_dict, conns_list, feature_function, to_file):
print("\tExtract features: [..]", end=' ')
example_list = []
for DocID, sent_index, conn_indices in conns_list:
feature = feature_function(parse_dict, DocID, sent_index, conn_indices)
example = Example("", feature)
example_list.append(example)
# write example_list to file
util.write_example_list_to_file(example_list, to_file)
print("\r\tExtract features: [OK]")
def arg_position_read_model_output(arg_position_model_output, conns_list):
SS_conns_list = []
PS_conns_list = []
# ['SS', 'PS'...]
pred_list = mallet_util.get_mallet_predicted_list(arg_position_model_output)
for pred, conn in zip(pred_list, conns_list):
if config.LABEL_TO_ARG_POSITION[pred] == "SS":
SS_conns_list.append(conn)
if config.LABEL_TO_ARG_POSITION[pred] == "PS":
PS_conns_list.append(conn)
return SS_conns_list, PS_conns_list
#[5, 6]
def divide_SS_conns_list(SS_conns_list):
SS_conns_parallel_list = []
SS_conns_not_parallel_list = []
for conn in SS_conns_list:
DocID, sent_index, conn_indices = conn
parallel = False
if len(conn_indices) > 1:
for i in range(len(conn_indices)):
if i + 1 < len(conn_indices) and conn_indices[i+1] - conn_indices[i] > 1:
parallel = True
if parallel:
SS_conns_parallel_list.append(conn)
else:
SS_conns_not_parallel_list.append(conn)
return SS_conns_parallel_list, SS_conns_not_parallel_list
def get_all_connectives_for_NT(parse_dict, conns_list):
connectives = []
for index, conn in enumerate(conns_list):
# turn to connective object
DocID, sent_index, conn_indices = conn
conn_name = get_conn_name(parse_dict, DocID, sent_index, conn_indices)
connective = Connective(DocID, sent_index, conn_indices, conn_name)
connective.relation_ID = index
connectives.append(connective)
return connectives
def constituent_print_feature(parse_dict, connectives, feature_function, to_file):
print("\tExtract features: [..]", end=' ')
example_list = []
# total = float(len(connectives))
for curr_index, connective in enumerate(connectives):
# print "process: %.2f%%.\r" % ((curr_index + 1)/total*100),
constituents = _get_constituents(parse_dict, connective)
constituents = sorted(constituents, key=lambda constituent: constituent.indices[0]) # sort by age
# extract features for each constituent
for i, constituent in enumerate(constituents):
feature = feature_function(parse_dict, constituent, i, constituents)
example = Example("", feature)
example.comment = "%s|%s" % (constituent.connective.relation_ID, " ".join([str(t) for t in constituent.get_indices()]))
example_list.append(example)
if constituents == []:
feature = Feature("", 0, {})
example = Example("", feature)
example.comment = "%s|%s" % (curr_index, " ")
example_list.append(example)
# write example_list to file
util.write_example_list_to_file(example_list, to_file)
print("\r\tExtract features: [OK]")
def constituent_read_model_output(
constituent_feat_path, constituent_model_output, parse_dict, conns_list):
feat_file = open(constituent_feat_path)
pred_list = mallet_util.get_mallet_predicted_list(constituent_model_output)
feature_list = [line.strip() for line in feat_file]
# relation_dict[relation_ID] = {(['1', '2'],'Arg1')....}
relation_dict = {}
for feature_line, predicted in zip(feature_list, pred_list):
comment = feature_line.split("#")[1].strip()
relation_ID = int(comment.split("|")[0].strip())
constituent_indices = comment.split("|")[1].strip().split(" ")
if relation_ID not in relation_dict:
relation_dict[relation_ID] = [(constituent_indices, predicted)]
else:
relation_dict[relation_ID].append((constituent_indices, predicted))
# merge arg1(arg2) for each relation
# relation_dict[relation_ID] = ([0,1],[2,3])
for relation_ID in list(relation_dict.keys()):
list_ = relation_dict[relation_ID]
Arg1_list = []
Arg2_list = []
for span, label in list_:
if label == "Arg1":
Arg1_list.extend(span)
if label == "Arg2":
Arg2_list.extend(span)
Arg1_list = sorted([int(item) for item in Arg1_list])
Arg2_list = sorted([int(item) for item in Arg2_list])
relation_dict[relation_ID] = (Arg1_list, Arg2_list)
temp = []
source = "SS"
for i, conn in enumerate(conns_list):
DocID, sent_index, conn_indices = conn
Arg1_list, Arg2_list = relation_dict[i]
Arg1_list = merge_NT_Arg(Arg1_list, parse_dict, DocID, sent_index)
Arg2_list = merge_NT_Arg(Arg2_list, parse_dict, DocID, sent_index)
if Arg1_list != [] and Arg2_list != []:
temp.append((source, DocID, sent_index, conn_indices, Arg1_list, Arg2_list))
else:
pass
# Arg1 or Arg2 is not identified
# temp.append((source, DocID, sent_index, conn_indices, Arg1_list, Arg2_list))
# if Arg1_list == []:
# print "Arg1###" + DocID, sent_index, conn_indices
# if Arg2_list == []:
# print "Arg2###" + DocID, sent_index, conn_indices
# if Arg1_list == [] and Arg2_list == []:
# print "Both###" + DocID, sent_index, conn_indices
return temp
def merge_NT_Arg(Arg_list, parse_dict, DocID, sent_index):
punctuation = """!"#&'*+,-..../:;<=>?@[\]^_`|~""" + "``" + "''"
if len(Arg_list) <= 1:
return Arg_list
temp = []
# scan the missing parts, if it is the punctuation, then make up
for i, item in enumerate(Arg_list):
if i <= len(Arg_list) - 2:
temp.append(item)
next_item = Arg_list[i + 1]
if next_item - item > 1:
flag = 1
for j in range(item + 1, next_item):
if parse_dict[DocID]["sentences"][sent_index]["words"][j][0] not in punctuation:
flag = 0
break
if flag == 1:# make up
temp += list(range(item + 1, next_item))
temp.append(Arg_list[-1])
Arg = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) for index in temp]
# remove the leading or tailing punctuations
Arg = util.list_strip_punctuation(Arg)
Arg = [item[0] for item in Arg]
return Arg
def get_Args_for_SS_parallel_conns(parse_dict, SS_conns_parallel_list):
temp = []
source = "SS"
for conn in SS_conns_parallel_list:
DocID, sent_index, conn_indices = conn
if len(conn_indices) == 2:# if then ,either or, neither nor
conn_1_index = conn_indices[0]
conn_2_index = conn_indices[1]
Arg1 = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) \
for index in range(conn_1_index+1, conn_2_index)]
sent_length = len(parse_dict[DocID]["sentences"][sent_index]["words"])
Arg2 = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) \
for index in range(conn_2_index+1, sent_length)]
Arg1 = util.list_strip_punctuation(Arg1)
Arg2 = util.list_strip_punctuation(Arg2)
Arg1 = [item[0] for item in Arg1]
Arg2 = [item[0] for item in Arg2]
temp.append((source, DocID, sent_index, conn_indices, Arg1, Arg2))
elif len(conn_indices) == 8:# on the one hand on the other hand
conn_1_index = conn_indices[3]
conn_2_index = conn_indices[7]
Arg1 = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) \
for index in range(conn_1_index+1, conn_2_index)]
sent_length = len(parse_dict[DocID]["sentences"][sent_index]["words"])
Arg2 = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) \
for index in range(conn_2_index+1, sent_length)]
Arg1 = util.list_strip_punctuation(Arg1)
Arg2 = util.list_strip_punctuation(Arg2)
Arg1 = [item[0] for item in Arg1]
Arg2 = [item[0] for item in Arg2]
temp.append((source, DocID, sent_index, conn_indices, Arg1, Arg2))
return temp
def get_Args_for_PS_conns(parse_dict, PS_conns_list):
source = "PS"
temp = []
for conn in PS_conns_list:
DocID, sent_index, conn_indices = conn
if sent_index - 1 < 0:
continue
# the length of the previous sentence
prev_length = len(parse_dict[DocID]["sentences"][sent_index - 1]["words"])
Arg1 = [(index, parse_dict[DocID]["sentences"][sent_index - 1]["words"][index][0])
for index in range(0, prev_length)]
Arg1 = util.list_strip_punctuation(Arg1)
# the length of the current sentence
curr_length = len(parse_dict[DocID]["sentences"][sent_index]["words"])
Arg2 = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) for index in range(0, curr_length)]
Arg2 = util.list_strip_punctuation(Arg2)
Arg1 = [item[0] for item in Arg1]
Arg2 = [item[0] for item in Arg2]
temp.append((source, DocID, sent_index, conn_indices, Arg1, Arg2))
return temp
def explicit_clf_print_feature(parse_dict, conns_list_args, feature_function, to_file):
print("\tExtract features: [..]", end=' ')
example_list = []
for conn in conns_list_args:
source, DocID, sent_index, conn_indices, Arg1, Arg2 = conn
connective = Connective(DocID, sent_index, conn_indices, "")
feature = feature_function(parse_dict, connective)
example = Example("", feature)
example_list.append(example)
# write example_list to file
util.write_example_list_to_file(example_list, to_file)
print("\r\tExtract features: [OK]")
def explicit_clf_read_model_output(explicit_model_output, conns_list_args):
pred_list = mallet_util.get_mallet_predicted_list(explicit_model_output)
pred_list = [config.Label_To_Sense[item] for item in pred_list]
temp = []
for pred, conn in zip(pred_list, conns_list_args):
source, DocID, sent_index, conn_indices, Arg1, Arg2 = conn
temp.append((source, DocID, sent_index, conn_indices, Arg1, Arg2, pred))
return temp
def get_explicit_relations(parse_dict, conns_args_sense_list):
SS_explicit_relations = []
PS_explicit_relations = []
for index, conn in enumerate(conns_args_sense_list):
source, DocID, sent_index, conn_indices, Arg1, Arg2, sense = conn
if source == "SS":
conn_token_list = get_doc_offset(parse_dict, DocID, sent_index, conn_indices)
Arg1_list = get_doc_offset(parse_dict, DocID, sent_index, Arg1)
Arg2_list = get_doc_offset(parse_dict, DocID, sent_index, Arg2)
# print Arg1
# print Arg1_list
relation = {}
relation["ID"] = index
relation['DocID'] = DocID
relation['Arg1'] = {}
relation['Arg1']['TokenList'] = Arg1_list
relation['Arg2'] = {}
relation['Arg2']['TokenList'] = Arg2_list
relation['Type'] = 'Explicit'
relation['Sense'] = [sense]
relation['Connective'] = {}
relation['Connective']['TokenList'] = conn_token_list
# add four attributes: Arg1_sent_index, Arg2_sent_index, conn_name, conn_sent_offset
relation["Arg1_sent_index"] = sent_index
relation["Arg2_sent_index"] = sent_index
relation["conn_sent_offset"] = conn_indices
relation["conn_name"] = get_conn_name(parse_dict, DocID, sent_index, conn_indices)
SS_explicit_relations.append(relation)
if source == "PS":
if sent_index - 1 < 0:
continue
conn_token_list = get_doc_offset(parse_dict, DocID, sent_index, conn_indices)
Arg1_list = get_doc_offset(parse_dict, DocID, sent_index - 1, Arg1)
Arg2_list = get_doc_offset(parse_dict, DocID, sent_index, Arg2)
relation = {}
relation["ID"] = index
relation['DocID'] = DocID
relation['Arg1'] = {}
relation['Arg1']['TokenList'] = Arg1_list
relation['Arg2'] = {}
relation['Arg2']['TokenList'] = Arg2_list
relation['Type'] = 'Explicit'
relation['Sense'] = [sense]
relation['Connective'] = {}
relation['Connective']['TokenList'] = conn_token_list
# add four attributes: Arg1_sent_index, Arg2_sent_index, conn_name, conn_sent_offset
relation["Arg1_sent_index"] = sent_index - 1
relation["Arg2_sent_index"] = sent_index
relation["conn_sent_offset"] = conn_indices
relation["conn_name"] = get_conn_name(parse_dict, DocID, sent_index, conn_indices)
PS_explicit_relations.append(relation)
return SS_explicit_relations + PS_explicit_relations
def test_explicit_relations(explicit_relations):
output = open(config.PARSER_EXPLICIT_REATION_PATH, 'w')
for relation in explicit_relations:
output.write('%s\n' % json.dumps(relation))
output.close()
print("-" * 120 + "\n Explicit Relation \n" + "-" * 120)
cmd = "python "+config.SCORER_PATH+" " \
" "+config.JSON_GOLD_EXPLICIT_PATH+" "+config.PARSER_EXPLICIT_REATION_PATH+" "
os.system(cmd)
def test_non_explicit_relations(non_explicit_relations):
output = open(config.PARSER_NON_EXPLICIT_REATION_PATH, 'w')
for relation in non_explicit_relations:
output.write('%s\n' % json.dumps(relation))
output.close()
print("-" * 120 + "\n Non-Explicit Relation \n" + "-" * 120)
cmd = "python "+config.SCORER_PATH+" " \
" "+config.JSON_GOLD_NON_EXPLICIT_PATH+" "+config.PARSER_NON_EXPLICIT_REATION_PATH+" "
os.system(cmd)
def test_relation(relations):
output = open(config.PARSER_REATION_PATH, 'w')
for relation in relations:
output.write('%s\n' % json.dumps(relation))
output.close()
print("-" * 120 + "\n All Relation \n" + "-" * 120)
cmd = "python "+config.SCORER_PATH+" " \
" "+config.PDTB_ORIGIN_DEV_PATH+" "+config.PARSER_REATION_PATH+" "
os.system(cmd)
def get_adjacent_non_exp_list(parse_dict, PS_conns_list):
exp_rel_sent_pairs = {}# [DocID] = [(1,2),(8,9)...]
for conn in PS_conns_list:
DocID, sent_index, conn_indices = conn
if sent_index == 0:
continue
if DocID not in exp_rel_sent_pairs:
exp_rel_sent_pairs[DocID] = [(sent_index - 1, sent_index)]
else:
exp_rel_sent_pairs[DocID].append((sent_index - 1, sent_index))
for DocID in exp_rel_sent_pairs:
exp_rel_sent_pairs[DocID] = set(exp_rel_sent_pairs[DocID])
#[(DocID,sent1_index,sent2_index) ]
adjacent_non_exp_list = []
for DocID in parse_dict:
sent_count = len(parse_dict[DocID]["sentences"])
adj_pair_set = _get_adj_pair_set(sent_count)
adj_exp_pair_set = set([])
if DocID in exp_rel_sent_pairs:
adj_exp_pair_set = exp_rel_sent_pairs[DocID]
adj_non_exp_pair_set = adj_pair_set - adj_exp_pair_set
for sent1_index, sent2_index in adj_non_exp_pair_set:
adjacent_non_exp_list.append((DocID, sent1_index, sent2_index))
# remove inter-paragraph sentence pairs
adjacent_non_exp_list = _remove_inter_paragraph_sent_pairs(parse_dict, adjacent_non_exp_list)
return adjacent_non_exp_list
#[(DocID,sent1_index,sent2_index) ]
# remove inter-paragraph sentence pairs
def _remove_inter_paragraph_sent_pairs(parse_dict, adjacent_non_exp_list):
temp = []
for DocID, sent1_index, sent2_index in adjacent_non_exp_list:
p1 = parse_dict[DocID]["sentences"][sent1_index]["paragraph"]
p2 = parse_dict[DocID]["sentences"][sent2_index]["paragraph"]
if p1 < 0 or p2 < 0:# can not get paragraph info
temp.append((DocID, sent1_index, sent2_index))
continue
if p1 == p2:# in same paragraph
temp.append((DocID, sent1_index, sent2_index))
return temp
#[(DocID,sent1_index,sent2_index) ]
def get_non_explicit_relations(parse_dict, adjacent_non_exp_list):
non_explicit_relations = []
for index, (DocID, sent1_index, sent2_index) in enumerate(adjacent_non_exp_list):
Arg1_offset_in_sent = _non_explicit_Arg_offset_in_sent(parse_dict, DocID, sent1_index)
Arg2_offset_in_sent = _non_explicit_Arg_offset_in_sent(parse_dict, DocID, sent2_index)
Arg1_TokenList = [ [-1, -1, -1, sent1_index, offset] for offset in Arg1_offset_in_sent]
Arg2_TokenList = [ [-1, -1, -1, sent2_index, offset] for offset in Arg2_offset_in_sent]
relation = {}
relation["ID"] = index
relation['DocID'] = DocID
relation['Arg1'] = {}
relation['Arg1']['TokenList'] = Arg1_TokenList
relation['Arg2'] = {}
relation['Arg2']['TokenList'] = Arg2_TokenList
relation['Type'] = 'Implicit'
relation['Connective'] = {}
relation['Connective']['TokenList'] = []
non_explicit_relations.append(relation)
return non_explicit_relations
def _non_explicit_Arg_offset_in_sent(parse_dict, DocID, sent_index):
curr_length = len(parse_dict[DocID]["sentences"][sent_index]["words"])
Arg = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0])
for index in range(0, curr_length)]
Arg = util.list_strip_punctuation(Arg)
Arg = [item[0] for item in Arg]
return Arg
def non_explicit_clf_print_feature(parse_dict, non_explicit_relations, feature_function, non_explicit_context_dict, prev_context_conn, to_file):
print("\tExtract features: [..]", end=' ')
example_list = []
for relation in non_explicit_relations:
feature_1 = feature_function(relation, parse_dict)
feature_2 = prev_context_conn(relation, parse_dict, non_explicit_context_dict)
feature = util.mergeFeatures([feature_1, feature_2])
example = Example("", feature)
example_list.append(example)
# write example_list to file
util.write_example_list_to_file(example_list, to_file)
print("\r\tExtract features: [OK]")
def non_explicit_read_model_output(non_explicit_model_output, parse_dict, non_explicit_relations):
pred_list = mallet_util.get_mallet_predicted_list(non_explicit_model_output)
pred_list = [config.Label_To_Sense[item] for item in pred_list]
temp = []
for sense, relation in zip(pred_list, non_explicit_relations):
relation['Sense'] = [sense]
temp.append(relation)
return temp
def cnn_non_explicit_read_model_output(cnn_output,non_explicit_relations):
import numpy as np
pred_list = [cnn_config.Label_To_Sense[np.argmax(item)] for item in cnn_output]
temp = []
for sense, relation in zip(pred_list, non_explicit_relations):
relation['Sense'] = [sense]
temp.append(relation)
return temp
def divide_non_explicit_relations(non_explicit_relations, parse_dict):
EntRel_relations = []
Implicit_AltLex_relations = []
for relation in non_explicit_relations:
if relation['Sense'][0] == "EntRel":#= "EntRel"
DocID = relation["DocID"]
Arg1_offset_in_sent = [item[4] for item in relation["Arg1"]["TokenList"]]
Arg2_offset_in_sent = [item[4] for item in relation["Arg2"]["TokenList"]]
Arg1_sent_index = relation["Arg1"]["TokenList"][0][3]
Arg2_sent_index = relation["Arg2"]["TokenList"][0][3]
Arg1_list = get_doc_offset(parse_dict, DocID, Arg1_sent_index, Arg1_offset_in_sent)
Arg2_list = get_doc_offset(parse_dict, DocID, Arg2_sent_index, Arg2_offset_in_sent)
relation['Arg1']['TokenList'] = Arg1_list
relation['Arg2']['TokenList'] = Arg2_list
EntRel_relations.append(relation)
else:
Implicit_AltLex_relations.append(relation)
# print "EntRel_relations:" + str(len(EntRel_relations))
# print "Implicit_AltLex_relations:" + str(len(Implicit_AltLex_relations))
return EntRel_relations, Implicit_AltLex_relations
def attri_print_feature(parse_dict, relations, attribution_feat_func, to_file):
example_list = []
total = float(len(relations))
for curr_index, relation in enumerate(relations):
sys.stdout.flush()
print("Extract Attribution Feature: %.2f%%.\r" % ((curr_index + 1)/total*100), end=' ')
for arg_clauses in _get_arg_clauses(parse_dict, relation):
if arg_clauses == []: continue
for clause_index in range(len(arg_clauses.clauses)):
feature = attribution_feat_func(arg_clauses, clause_index, parse_dict)
#example
example = Example("", feature)
example.comment = "%s|%s|%s" % \
(arg_clauses.relation_ID, arg_clauses.Arg, " ".join([str(i) for i in arg_clauses.clauses[clause_index][0]]))
example_list.append(example)
# write example_list to file
util.write_example_list_to_file(example_list, to_file)
print("Attribution Feature : Done!")
def attri_read_model_output(attribution_feat_path, attribution_model_output, parse_dict, non_explicit_relations):
feat_file = open(attribution_feat_path)
pred_list = mallet_util.get_mallet_predicted_list(attribution_model_output)
implicit_relations = {}
for relation in non_explicit_relations:
relation_ID = relation["ID"]
DocID = relation["DocID"]
Arg1_sent_indices = sorted([item[3] for item in relation["Arg1"]["TokenList"]])
Arg2_sent_indices = sorted([item[3] for item in relation["Arg2"]["TokenList"]])
sent1_index = Arg1_sent_indices[0]
sent2_index = Arg2_sent_indices[0]
Arg1_offset_in_sent = [item[4] for item in relation["Arg1"]["TokenList"]]
Arg2_offset_in_sent = [item[4] for item in relation["Arg2"]["TokenList"]]
Arg1 = [(index, parse_dict[DocID]["sentences"][sent1_index]["words"][index][0]) for index in Arg1_offset_in_sent]
Arg2 = [(index, parse_dict[DocID]["sentences"][sent2_index]["words"][index][0]) for index in Arg2_offset_in_sent]
Arg1 = util.list_strip_punctuation(Arg1)
Arg2 = util.list_strip_punctuation(Arg2)
implicit_relations[(relation_ID, "Arg1")] = Arg1
implicit_relations[(relation_ID, "Arg2")] = Arg2
feature_list = [line.strip() for line in feat_file]
# relation_dict[(relation_ID,Arg)] = {([1, 2],'yes')....}
relation_dict = {}
for feature_line, predicted in zip(feature_list, pred_list):
comment = feature_line.split("#")[1].strip()
relation_ID = int(comment.split("|")[0].strip())
Arg = comment.split("|")[1].strip()
attri_indices = [int(i) for i in comment.split("|")[2].strip().split(" ")]
if (relation_ID, Arg) not in relation_dict:
relation_dict[(relation_ID, Arg)] = [(attri_indices, predicted)]
else:
relation_dict[(relation_ID, Arg)].append((attri_indices, predicted))
# remove the attribution part for each argument of the relation
for (relation_ID, Arg) in list(relation_dict.keys()):
list = relation_dict[(relation_ID, Arg)]#[([1, 2],'yes')....]
for span, label in list:
if label == "yes":
implicit_Arg = implicit_relations[(relation_ID, Arg)]#dict[(relation_ID,Arg)] = [(1,"I")..]
part1 = []
part2 = []
flag = 0
for index, word in implicit_Arg:
if flag == 0 and index not in span:
part1.append((index, word))
if index in span:
flag = 1
if flag == 1 and index not in span:
part2.append((index, word))
implicit_relations[(relation_ID, Arg)] = util.list_strip_punctuation(part1) + util.list_strip_punctuation(part2)
arg_list = [item[0] for item in implicit_relations[(relation_ID, Arg)]]
if arg_list == []:
if Arg == "Arg1":
relation_dict[(relation_ID, Arg)] = list[-1][0]
else:
relation_dict[(relation_ID, Arg)] = list[0][0]
else:
relation_dict[(relation_ID, Arg)] = arg_list
temp = []
for relation in non_explicit_relations:
relation_ID = relation["ID"]
DocID = relation["DocID"]
if (relation_ID, "Arg1") not in relation_dict:
print("11")
Arg1_offset_in_sent = [item[4] for item in relation["Arg1"]["TokenList"]]
else:
Arg1_offset_in_sent = relation_dict[(relation_ID, "Arg1")]
if (relation_ID, "Arg2") not in relation_dict:
print("22")
Arg2_offset_in_sent = [item[4] for item in relation["Arg2"]["TokenList"]]
else:
Arg2_offset_in_sent = relation_dict[(relation_ID, "Arg2")]
# Arg2_offset_in_sent = [item[4] for item in relation["Arg2"]["TokenList"]]
Arg1_sent_index = relation["Arg1"]["TokenList"][0][3]
Arg2_sent_index = relation["Arg2"]["TokenList"][0][3]
Arg1_list = get_doc_offset(parse_dict, DocID, Arg1_sent_index, Arg1_offset_in_sent)
Arg2_list = get_doc_offset(parse_dict, DocID, Arg2_sent_index, Arg2_offset_in_sent)
relation['Arg1']['TokenList'] = Arg1_list
relation['Arg2']['TokenList'] = Arg2_list
curr_length_1 = len(parse_dict[DocID]["sentences"][Arg1_sent_index]["words"])
Arg1_sent_text = [parse_dict[DocID]["sentences"][Arg1_sent_index]["words"][index][0] for index in range(0, curr_length_1)]
Arg1_text = [parse_dict[DocID]["sentences"][Arg1_sent_index]["words"][index][0] for index in Arg1_offset_in_sent]
# print DocID, Arg1_sent_index
# print " ".join(Arg1_sent_text)
# print " ".join(Arg1_text)
temp.append(relation)
return temp
def _get_arg_clauses(parse_dict, relation):
return [_arg_clauses(parse_dict, relation, "Arg1"), _arg_clauses(parse_dict, relation, "Arg2")]
def _get_arg1_clauses(parse_dict, relation):
return [_arg_clauses(parse_dict, relation, "Arg1")]
def _get_arg2_clauses(parse_dict, relation):
return [_arg_clauses(parse_dict, relation, "Arg2")]
def _arg_clauses(parse_dict, relation, Arg):
DocID = relation["DocID"]
Arg_sent_indices = sorted([item[3] for item in relation[Arg]["TokenList"]])
Arg_token_indices = sorted([item[4] for item in relation[Arg]["TokenList"]])
if len(set(Arg_sent_indices)) != 1:
return []
relation_ID = relation["ID"]
sent_index = Arg_sent_indices[0]
sent_tokens = [(index, parse_dict[DocID]["sentences"][sent_index]["words"][index][0]) for index in Arg_token_indices]
punctuation = "...,:;?!~--"
# first, use punctuation symbols to split the sentence
_clause_indices_list = []#[[(1,"I")..], ..]
temp = []
for index, word in sent_tokens:
if word not in punctuation:
temp.append((index, word))
else:
if temp != []:
_clause_indices_list.append(temp)
temp = []
if temp != []:
_clause_indices_list.append(temp)
clause_indices_list = []
for clause_indices in _clause_indices_list:
temp = util.list_strip_punctuation(clause_indices)
if temp != []:
clause_indices_list.append([item[0] for item in temp])
# then use SBAR tag in its parse tree to split each part into clauses.
parse_tree = parse_dict[DocID]["sentences"][sent_index]["parsetree"].strip()
syntax_tree = Syntax_tree(parse_tree)
if syntax_tree.tree == None:
return []
clause_list = []
for clause_indices in clause_indices_list:
clause_tree = _get_subtree(syntax_tree, clause_indices)
# BFS，
flag = 0
for node in clause_tree.tree.traverse(strategy="levelorder"):
if node.name == "SBAR":
temp1 = [node.index for node in node.get_leaves()]
temp2 = sorted(list(set(clause_indices) - set(temp1)))
if temp2 == []:
clause_list.append(temp1)
else:
if temp1[0] < temp2 [0]:
clause_list.append(temp1)
clause_list.append(temp2)
else:
clause_list.append(temp2)
clause_list.append(temp1)
flag = 1
break
if flag == 0:
clause_list.append(clause_indices)
clauses = []# [([1,2,3],yes), ([4, 5],no), ]
for clause_indices in clause_list:
clauses.append((clause_indices, ""))
# print DocID, sent_index
# print " ".join([parse_dict[DocID]["sentences"][sent_index]["words"][index][0] for index in Arg_token_indices])
# print clauses
return Arg_Clauses(relation_ID, Arg, DocID, sent_index, clauses)
def _get_subtree(syntax_tree, clause_indices):
copy_tree = copy.deepcopy(syntax_tree)
for index, leaf in enumerate(copy_tree.tree.get_leaves()):
leaf.add_feature("index",index)
clause_nodes = []
for index in clause_indices:
node = copy_tree.get_leaf_node_by_token_index(index)
clause_nodes.append(node)
for node in copy_tree.tree.traverse(strategy="levelorder"):
node_leaves = node.get_leaves()
if set(node_leaves) & set(clause_nodes) == set([]):
node.detach()
return copy_tree
# [(0, 1), (1, 2), (2, 3), (3, 4)]
def _get_adj_pair_set(length):
i = 0
list = []
while i < length -1:
list.append((i, i+1))
i += 1
return set(list)
def get_doc_offset(parse_dict, DocID, sent_index, list):
offset = 0
for i in range(sent_index):
offset += len(parse_dict[DocID]["sentences"][i]["words"])
temp = []
for item in list:
temp.append(item + offset)
return temp
def _get_constituents(parse_dict, connective):
DocID = connective.DocID
sent_index = connective.sent_index
parse_tree = parse_dict[DocID]["sentences"][sent_index]["parsetree"].strip()
syntax_tree = Syntax_tree(parse_tree)
if syntax_tree.tree == None:
return []
conn_indices = connective.token_indices
constituent_nodes = []
if len(conn_indices) == 1:# like and or so...
conn_node = syntax_tree.get_leaf_node_by_token_index(conn_indices[0]).up
else:
conn_node = syntax_tree.get_common_ancestor_by_token_indices(conn_indices)
conn_leaves = set([syntax_tree.get_leaf_node_by_token_index(conn_index) for conn_index in conn_indices])
children = conn_node.get_children()
for child in children:
leaves = set(child.get_leaves())
if conn_leaves & leaves == set([]):
constituent_nodes.append(child)
curr = conn_node
while not curr.is_root():
constituent_nodes.extend(syntax_tree.get_siblings(curr))
curr = curr.up
# obtain the Constituent object according to the node.
constituents = []
for node in constituent_nodes:
cons = Constituent(syntax_tree, node)
cons.connective = connective
constituents.append(cons)
return constituents
def get_conn_name(parse_dict, DocID, sent_index, conn_indices):
# obtain the name of the connective
conn_name = " ".join([parse_dict[DocID]["sentences"][sent_index]["words"][word_token][0] \
for word_token in conn_indices ])
return conn_name.lower()
import codecs
def add_paragraph_info_for_parse(parse_dict, raw_path):
for DocID in parse_dict:
try:
# raw_file = open("%s/%s" % (raw_path, DocID))
raw_file = codecs.open("%s/%s" % (raw_path, DocID), encoding="utf-8", errors="ignore")
paragTexts = getParagTexts(raw_file)# ["IamaDoy","asas']
for sent_index in range(len(parse_dict[DocID]["sentences"])):
sent_words_list = [word[0] for word in parse_dict[DocID]["sentences"][sent_index]["words"]]
ParagIndex = getParagIndex(paragTexts, sent_words_list)
parse_dict[DocID]["sentences"][sent_index]["paragraph"] = ParagIndex
except IOError:
# if failed, set parse_dict[DocID]["sentences"][sent_index]["paragraph"] = -1
for sent_index in range(len(parse_dict[DocID]["sentences"])):
parse_dict[DocID]["sentences"][sent_index]["paragraph"] = -1
def add_paragraph_info(raw_file, doc):
paragTexts = getParagTexts(raw_file)
for sentence in doc["sentences"]:
sent_words_list = [word[0] for word in sentence["words"]]
ParagIndex = getParagIndex(paragTexts, sent_words_list)
sentence["paragraph"] = ParagIndex
# print sent_words_list
# print ParagIndex
pass
def getParagTexts(raw_file):
text = [line.strip() for line in raw_file.readlines()]+['']
t = 0
for line in text:
if line == '.START' or line == "":
t += 1
else:
break
text = text[t:]
paragTexts =[]
paragText = ""
for line in text:
if line != '':
# line = str(line, "utf-8", errors='ignore')
paragText += line.replace(" ", "")
else:
if paragText != "":
paragText = util.removePuctuation(paragText)
paragTexts.append(paragText)
paragText = ""
return paragTexts
def getParagIndex(paragTexts, sent_tokens):
paragIndex = -1
sent = "".join(sent_tokens)
sent = sent.replace("-LCB-", "")
sent = sent.replace("-LRB-", "")
sent = sent.replace("-RCB-", "")
sent = sent.replace("-RRB-", "")
sent = util.removePuctuation(sent)
matchedParag = set([])
for index, paragText in enumerate(paragTexts):
if sent in paragText:
matchedParag.add(index)
#matchedParag might be [1],[3,6] , take the minimum
if matchedParag != set([]):
paragIndex = min(matchedParag)
# remove the sent which have been matched
paragTexts[paragIndex] = paragTexts[paragIndex].replace(sent, "", 1)
# if paragIndex == -1:
# print sent_tokens
# raise ValueError("sentence : '%s' , can not get the paragIndex" % (sent) )
return paragIndex
def put_feature_to_model(feature_path, model_path, model_output_path):
cmd = config.MALLET_PATH + "/bin/mallet classify-file --input " + feature_path + " --output " + model_output_path + " --classifier " + model_path
os.system(cmd)
def _change_feature_dimension(test_file_path, n_features):
file = open(test_file_path)
lines = []
flag = 0
for line in file:# 175:1 21381:1 #
line = line.rstrip()
if flag == 0 and line.split("#")[0].strip() != "":
last_feat_dimension = int(line.split("#")[0].rstrip().split(" ")[-1].split(":")[0])
if last_feat_dimension < n_features:
line = line.split("#")[0] +"%d:0 #" % n_features + line.split("#")[1]
flag = 1
line = "-1" + line
lines.append(line)