-
Notifications
You must be signed in to change notification settings - Fork 0
/
parser.py
1515 lines (1319 loc) · 57.1 KB
/
parser.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
#!/usr/bin/env python3
"""
TODO: create a LinkFeature superclass for parsing the linked features to avoid code duplication and increase efficiency
TODO: refactor current Document class into a Document class and an Annotation class:
Document contains the text, sentences, tokens, metadata and annotation_documents.
Annotation contains all annotation_documents and the parsing code.
This allows for less duplication (text, tokens, sentences and metadata are all the same for the same doc).
TODO: write generic parser for webanno features (Link, Slot, etc):
cassis is slower than this prototype but more general
parser.py
sentivent_webannoparser
10/4/18
Copyright (c) Gilles Jacobs. All rights reserved.
"""
from __future__ import annotations
import sys
sys.path.append("/home/gilles/repos/")
import settings
from dataclasses import dataclass, field
from zipfile import ZipFile
import xml.dom.minidom as md
# import cassis
import fnmatch
import spacy
from pathlib import Path
from typing import List, Any
from sentivent_webannoparser.util import flatten, count_avg, pickle_webanno_project
import os
import multiprocessing
import json
from itertools import groupby, combinations_with_replacement
from copy import copy
from math import log
@dataclass
class Element:
text: str
begin: int = field(repr=False)
end: int = field(repr=False)
element_id: int = field(repr=False)
annotator_id: str
document_title: str
in_document: Any = field(repr=False)
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
def friendly_id(self):
"""
Make a representation of object that is easy to find in corpus.
:return:
"""
id = f"{self.annotator_id}_{self.document_title.split('_')[0]}"
try: # try making an sentence identifier if there is an in_sentence attrib
sen_id = ",".join(str(se.element_id + 1) for se in self.in_sentence)
id += f"_s{sen_id}"
except Exception as e:
print(e)
pass
if isinstance(self, Event):
id += f"_{self.event_fulltype}"
elif isinstance(self, Participant) or isinstance(self, Filler):
id += f"_{self.role}"
text_ellips = (
(self.text[:15] + ".." + self.text[-15:])
if len(self.text) > 32
else self.text
)
id += f"-{text_ellips}"
return id
def get_processed_sentence(self):
"""
Return the Spacy processed sentences in which the element is positioned.
:return:
"""
if not hasattr(self, "in_sentence"):
raise AttributeError(f"{self} does not have attribute 'in_sentence'.")
try:
sen_ixs = [sen.element_id for sen in self.in_sentence]
except TypeError as e:
sen_ixs = [self.in_sentence.element_id]
sen_procs = [
s
for i, s in enumerate(self.in_document.sentences_processed)
if i in sen_ixs
]
return sen_procs
def get_processed_tokens(self, **kwargs):
if hasattr(
self, "get_extent_tokens"
): # for events and sentimentexpr which have custom token getters
token_origs = self.get_extent_tokens(**kwargs)
else:
token_origs = self.tokens
sen_procs = self.get_processed_sentence()
# some annotation can exceptionally span multiple sentences (actually not the case in practice)
# groupby sentence to retrieve token corresponding to sentence
sen_key = lambda x: x.in_sentence.element_id
token_origs_by_sen = [
list(g) for k, g in groupby(sorted(token_origs, key=sen_key), sen_key)
]
token_procs = []
for sen_tok_orig, sen_proc in zip(token_origs_by_sen, sen_procs):
sen_token_orig_ixs = [t.index_sentence for t in sen_tok_orig]
sen_token_procs = [
t_proc for i, t_proc in enumerate(sen_proc) if i in sen_token_orig_ixs
]
token_procs.extend(sen_token_procs)
return token_procs
def check_pronominal(self):
"""
Check if the annotation is pronominal by full parsed PoS tag.
All tokens should be pronominal (works best for anaphoric pronominal mentions as intended).
:return:
"""
pronom_tags = ["PRP", "PRP$", "WDT", "WP", "WP$"]
token_procs = self.get_processed_tokens()
all_pronom = all(
t.tag_ in pronom_tags for t in token_procs
) # True if all tokens are pronom_tags
# print(f"{' '.join(t.text + '.' + t.tag_ for t in token_procs)}: Pronominal = {all_pronom}")
return all_pronom
def replace_by_canonical_referent(self, cross_sentence=True):
"""
Replace annotation unit by its Canonical referent.
:param cross_sentence: Set to False: Do not replace by CanonicalReferent if the link crosses sentence boundary.
:return:
"""
replaced_count = 0
# replace the participant in their containers
def replace_in_containers(myobj, replacing):
containers = {
"event": "participants",
"sentiment_expression": "targets",
"sentence": "participants",
}
for c_name, attrib_n in containers.items():
c_name = f"in_{c_name}"
if hasattr(myobj, c_name):
for c in getattr(myobj, c_name):
to_replace_in = getattr(c, attrib_n)
to_replace_in.append(replacing)
if myobj in to_replace_in:
to_replace_in.remove(myobj)
to_replace_in.sort(key=lambda x: x.begin)
if self.canonical_referents and self.canonical_referents != "from_canonref":
# sometimes there are multiple canonrefs tagged
# this can be a) annotation mistake or
# b) multiple reference to a group, e.g. "all" refers to three companies.
for canonref in self.canonical_referents:
# check whether canonical referent is in same sentence
same_sentence = [s.element_id for s in self.in_sentence] == [
s.element_id for s in canonref.in_sentence
]
# always replace when cross_sentence is true
# if cross_sentence is False (disallowed) only replace when canonref is in same sentence
if cross_sentence or same_sentence:
# replace element
replacing_participant = Participant(
canonref.text,
canonref.begin,
canonref.end,
canonref.element_id,
canonref.annotator_id,
canonref.document_title,
canonref.in_document,
self.role,
"from_canonref",
self.link_id,
canonref.tokens,
)
replacing_participant.in_sentence = self.in_sentence
replacing_participant.in_document = self.in_document
if hasattr(self, "in_sentiment_expression"):
replacing_participant.in_sentiment_expression = (
self.in_sentiment_expression
)
replacing_participant.from_original_participant = copy(self)
print(
f"Replaced {self} with {canonref}. (in same sentence:{same_sentence})"
)
replace_in_containers(self, replacing_participant)
# replace on document
self.in_document.participants.append(replacing_participant)
if self in self.in_document.participants:
self.in_document.participants.remove(self)
replaced_count += 1
return replaced_count
@dataclass
class Filler(Element):
role: str
link_id: int = field(repr=False)
tokens: List = field(default=None, repr=False)
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
@dataclass
class DiscontiguousTrigger(Element):
link_id: int = field(repr=False)
tokens: List = field(default=None, repr=False)
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
@dataclass
class CanonicalReferent(Element):
pronom_id: int = field(repr=False)
referent_id: int
tokens: List = field(default=None, repr=False)
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
@dataclass
class Participant(Element):
role: str
canonical_referents: List[CanonicalReferent] = field(repr=False)
link_id: int = field(repr=False)
tokens: List = field(default=None, repr=False)
def get_extent_text(self):
return " ".join(t.text for t in sorted(list(set(self.tokens)), key=lambda x: x.index_sentence))
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
class Sentence:
"""
Class to represent a sentence
Variables:
id - number of this sentence in the text
begin - position of the sentence start in the original text
end - position of the sentecne end in the original text
frames - list of frames contained in this sentence
"""
def __init__(
self,
element_id,
begin,
end,
tokens,
events,
sentiment_expressions,
participants,
fillers,
canonical_referents,
targets,
):
self.element_id = element_id
self.begin = begin
self.end = end
self.tokens = tokens
self.events = events
self.sentiment_expressions = sentiment_expressions
self.participants = participants
self.fillers = fillers
self.canonical_referents = canonical_referents
self.targets = targets
def __eq__(self, other):
if self.begin == other.begin and self.end == other.end:
if [x for x in self.events if not x in other.events] == [] and [
x for x in other.events if not x in self.events
] == []:
return True
return False
def __str__(self):
return " ".join(str(t) for t in self.tokens)
def __repr__(self):
"""
Custom repr for easier debugging.
:return: repr string
"""
text = str(self)
text_ellips = (text[:31] + ".." + text[-31:]) if len(text) > 64 else text
return f"{self.element_id}. {text_ellips}..."
@dataclass
class Event(Element):
event_type: str
event_subtype: str
event_fulltype: str
discontiguous_triggers: List[DiscontiguousTrigger] = field(repr=False)
participants: List[Participant] = field(repr=False)
fillers: List[Filler] = field(repr=False)
polarity_negation: str = field(repr=False)
modality: str = field(repr=False)
realis: str = field(repr=False)
polarity_sentiment: str = field(repr=True)
polarity_sentiment_scoped: str = field(repr=False)
tokens: List = field(default=None, repr=False)
in_sentence: List[Sentence] = field(default=None, repr=False)
coordination: List[Event] = field(default=None, repr=False)
coreferents: List[Event] = field(default=None)
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
def get_coref_attrib(self, attrib):
if self.coreferents is not None:
coref_attribs = []
for coref in self.coreferents:
walk_attrib = coref.get_coref_attrib(attrib)
if walk_attrib is not None:
coref_attribs.append(walk_attrib)
attrib_val = getattr(coref, attrib)
if attrib_val is not None:
coref_attribs.append(attrib_val)
if coref_attribs:
return flatten(coref_attribs)
else:
return None
else:
return None
def get_sentence_text(self):
return " ".join(t.text for sen in self.in_sentence for t in sen.tokens)
def get_extent_tokens(self, extent=["discontiguous_triggers"], source_order=True):
"""
Get a list of token objects for the event extent.
The extent can be set to include discontiguous_triggers, participants, and/or fillers.
Setting this to an empty list will only return the original
In the definition of an event nugget we include all of these.
:param extent: a list of elements with which the event annotation span can be extended. Default: discontigious_triggers, allowed: ["discontiguous_triggers", "participants", "fillers"]
:param source_order: Maintain the word order of tokens in the source AnnotationDocument.
Setting to False preserves the order of the annotation process by the annotators.
This is an unreliable approximation of token span salience and cannot be recommended. Default: True
:return: List of tokens
"""
all_tokens = self.tokens.copy()
core_sen_idx = all_tokens[0].in_sentence.element_id # to ensure discont is in same sentence
for ext in extent:
if getattr(self, ext):
all_tokens.extend(t for x in getattr(self, ext) for t in x.tokens if t.in_sentence.element_id == core_sen_idx)
all_tokens = list(
set(all_tokens)
) # this is necessary because trigger and participant spans can overlap
if source_order: # return tokens in the order they appear in source text
all_tokens.sort(key=lambda x: x.begin)
return all_tokens
def get_extent_token_ids(self, **kwargs):
"""
Get a list of token ids for the event extent.
Relies on Event.get_extent_tokens() for fetching the tokens.
:param kwargs:
:return: a list of token ids of the event extent that are unique in the document.
"""
token_span = self.get_extent_tokens(**kwargs)
return [t.index for t in token_span]
def get_extent_text(
self,
extent=["discontiguous_triggers", "participants", "fillers"],
source_order=True,
):
return " ".join(
t.text
for t in self.get_extent_tokens(extent=extent, source_order=source_order)
)
def _fix_false_discont(self):
fixed = [self.tokens]
new_discont = self.discontiguous_triggers[:]
idc = self.trigger_parts_idc
parts = self.trigger_parts
for i in range(len(idc)-1):
if idc[i][1] + 1 >= idc[i+1][0]: # check if adjacent and merge
if parts[i] not in fixed:
fixed.append(parts[i])
if parts[i] in [ndc.tokens for ndc in new_discont]:
new_discont = [dc for dc in new_discont if dc.tokens != parts[i]]
if parts[i+1] not in fixed:
fixed.append(parts[i+1])
if parts[i+1] in [ndc.tokens for ndc in new_discont]:
new_discont = [dc for dc in new_discont if dc.tokens != parts[i+1]]
self.tokens = sorted(list(set(i for l in fixed for i in l)), key=lambda x: x.index_sentence) # flatten, dedupe, and sort
self.discontiguous_triggers = new_discont
self.trigger_parts = sorted([self.tokens] + [d.tokens for d in self.discontiguous_triggers], key=lambda x: x[0].index_sentence)
self.trigger_parts_idc = [(part[0].index_sentence, part[-1].index_sentence) for part in self.trigger_parts]
def _generate_conti_combos(self, max_dist):
conti_combo = [self.trigger_parts[i:j+1] for i,j in combinations_with_replacement(range(len(self.trigger_parts)),2)]
conti_combo = [c for c in conti_combo if c[-1][-1].index_sentence - c[0][0].index_sentence < max_dist]
conti_combo_idc = [(c[0][0].index_sentence, c[-1][-1].index_sentence) for c in conti_combo]
conti_combo_tokens = [self.in_sentence[0].tokens[i[0]:i[1]+1] for i in conti_combo_idc]
return conti_combo_tokens
def _score_content_tokens(self, tokens):
# extract preproc token span
doc = tokens[0].get_processed_sentence()[0]
span_orig = doc[
tokens[0].index_sentence : tokens[-1].index_sentence + 1
]
# generate and score candidates
score = 0
for t in span_orig:
parts_idc = set(t.index_sentence for p in self.trigger_parts for t in p)
if t.i in parts_idc:
if t.pos_ in ["ADJ", "VERB", "ADV", "NUM", "NOUN", "ADP"]:
score += 1.0
elif t.pos_ in ["AUX", "PROPN"]:
score += 0.5
# boost core annotation
# if t.i in set(t.index_sentence for t in self.tokens):
# score += 0.5
return score
def preprocess_trigger(self,
fix_false_discont=True,
make_continuous_max_dist=0,
truncate_to_len=False,
):
self.tokens_orig = self.tokens[:]
self.discontiguous_triggers_orig = self.discontiguous_triggers[:] if self.discontiguous_triggers else None
if self.discontiguous_triggers:
self.trigger_parts = sorted([self.tokens] + [d.tokens for d in self.discontiguous_triggers], key=lambda x: x[0].index_sentence)
else:
self.trigger_parts = [self.tokens]
self.trigger_parts_idc = [(part[0].index_sentence, part[-1].index_sentence) for part in self.trigger_parts]
if self.discontiguous_triggers and fix_false_discont: # no discontinuous preproc needed
# first fix annotation artifacts with adjacent discontinuous parts to make them continuous
self._fix_false_discont()
if self.discontiguous_triggers: # generate all combinations of continual discont parts
print(f"-----Making discont parts. continuous: [{' ... '.join(' '.join(t.text for t in p) for p in self.trigger_parts)}]")
conti_combos = self._generate_conti_combos(make_continuous_max_dist)
conti_combos_scores = [self._score_content_tokens(tokens) for tokens in conti_combos]
# score by contentfullness scorer and get contentfullness/length ratio
conti_combos_scores_scaled = [(log(score))/(1+log(len(combo))) if score else 0 for combo, score in zip(conti_combos, conti_combos_scores)]
max_i = conti_combos_scores_scaled.index(max(conti_combos_scores_scaled))
for i, (combo, score, s_score) in enumerate(zip(conti_combos, conti_combos_scores, conti_combos_scores_scaled)):
prefix = " -> " if i == max_i else " "
print(f"{prefix}{' '.join(t.text for t in combo)} |\tscore: {score}\tlen ratio: {s_score}")
self.tokens = conti_combos[max_i]
@dataclass
class SentimentExpression(Element):
polarity_sentiment: str = field(repr=True)
polarity_sentiment_scoped: str = field(repr=False)
uncertain: str = field(repr=False)
negated: str = field(repr=True)
targets: List = field(repr=False)
tokens: List[Token] = field(default=None, repr=False)
in_sentence: List[Sentence] = field(default=None, repr=False)
def __str__(self):
id = f"|{self.annotator_id[:3]}|{self.in_document.document_id}|s{self.in_sentence[0].element_id:02d}|"
se_text = f"{id} <{self.polarity_sentiment_scoped.upper()[:3]}> {self.get_extent_text()}"
spacing = len(se_text) * " "
tgt_text = [f"|s{t.in_sentence[0].element_id:02d}| {t.get_extent_text()}" for t in self.targets]
return f"{se_text} --> " + f"\n{spacing} └-> ".join(tgt_text)
def __hash__(self):
return hash((self.document_title, self.element_id, self.text, self.begin, self.end))
def get_extent_tokens(self, extent=[], source_order=True):
"""
Get a list of token objects for the sentiment expression extent.
The extent can be set to include targets.
Setting this to an empty list will only return the original tokens.
:param extent: a list of elements with which the event annotation span can be extended. Default: []
:param source_order: Maintain the word order of tokens in the source AnnotationDocument.
Setting to False preserves the order of the annotation process by the annotators.
This is an unreliable approximation of token span salience and cannot be recommended. Default: True
:return: List of tokens
"""
all_tokens = self.tokens.copy()
for ext in extent:
if getattr(self, ext):
all_tokens.extend(t for x in getattr(self, ext) for t in x.tokens)
all_tokens = list(
set(all_tokens)
) # this is necessary because trigger and participant spans can overlap
if source_order: # return tokens in the order they appear in source text
all_tokens.sort(key=lambda x: x.begin)
return all_tokens
def get_extent_token_ids(self, **kwargs):
tokens = self.get_extent_tokens(**kwargs)
return [f"{self.document_title.split('_')[0]}_{t.index}" for t in tokens]
def get_extent_text(self, **kwargs):
return " ".join(t.text for t in self.get_extent_tokens(**kwargs))
@dataclass
class Token(Element):
index: int
event_extent: List[Event] = field(repr=False)
participant_extent: List[Participant] = field(repr=False)
filler_extent: List[Filler] = field(repr=False)
canonical_referent_extent: List[CanonicalReferent] = field(repr=False)
discontiguous_trigger_extent: List[DiscontiguousTrigger] = field(repr=False)
sentiment_expression_extent: List[SentimentExpression] = field(repr=False)
target_extent: List[Participant] = field(repr=False)
def get_token_id(self):
"""
Set token id based on document id + token position in text
:return:
"""
return f"{self.document_title}_{self.index}"
def __hash__(self):
return hash((self.element_id, self.text, self.begin, self.end))
def __str__(self):
return self.text
class SourceDocument:
# TODO Finish this so NLP processing is done on only this shared object and annotation_documents are held in AnnotionDocument
def __init__(self, annotation_documents):
self.title = next(annotation_documents).title
self.text = next(annotation_documents).text
self.annotations = List[AnnotationDocument]
class AnnotationDocument:
"""
Class to represent a WebAnno Annotation in the XMI format
Arguments:
file_name - name of the XMI file which contains the annotation
Variables:
text - textual representation of the annotated text
tagset - attributes used to describe a frame
sentences - list of sentences this document consists of
"""
def __init__(self, xmi_content, path="", *args, **kwargs):
self.path = path
self.annotator_id = self.path.split("/")[-1].replace(".xmi", "")
self.file_id = self.path.split("/")[-2]
print(f"Parsing doc {self.path}")
dom = md.parseString(xmi_content)
self.text = dom.getElementsByTagName("cas:Sofa")[0].getAttribute("sofaString")
self.title = dom.getElementsByTagName("type2:DocumentMetaData")[0].getAttribute(
"documentTitle"
)
self.document_id = self.title.split("_")[0]
self.events = None
self.fillers = None
self.discontiguous_triggers = None
self.canonical_referents = None
self.participants = None
self.sentences = None
self.tokens = None
# stationary views of the XMI items
sentences_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("type4:Sentence")
]
tokens_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("type4:Token")
]
events_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:A_Event")
]
coref_event_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:CorefEvent")
]
participant_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:B_Participant")
]
participantlink_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:A_EventC_ParticipantLink")
]
pronomcanonref_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:PronomCanonRef")
]
filler_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:C_FILLER")
]
fillerlink_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:A_EventD_FILLERLink")
]
discontiguous_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:D_Discontiguous")
]
discontiguouslink_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:A_EventF_DiscontiguousLink")
]
sentiment_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:E_Sentiment")
]
targetentity_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:E_SentimentA_TargetLink")
]
targeteventlink_xmidata = [
self.__convertAttributes__(node)
for node in dom.getElementsByTagName("custom:E_SentimentB_TargetEventLink")
]
# Create canonical referents objects
canonical_referents = []
for pcr in pronomcanonref_xmidata:
canonical_referents.append(
CanonicalReferent(
*self.__extract_default_xmi(pcr),
self.annotator_id,
self.title,
self,
pcr["Governor"],
pcr["Dependent"],
)
)
if canonical_referents:
self.canonical_referents = canonical_referents
# Create participant objects from links, get element
participants = []
for part in participant_xmidata:
text, begin, end, element_id = self.__extract_default_xmi(part)
# parse all links that refer to the participants
# parse participant_links which link to roles in events
participant_links = list(
filter(
lambda part_link: int(part_link["target"]) == element_id,
participantlink_xmidata,
)
)
# parse canon referent
if self.canonical_referents:
canonref = list(
filter(
lambda canonref: str(canonref.pronom_id) == str(element_id),
self.canonical_referents,
)
)
else:
canonref = None
# participant spans can have multiple roles of different events but if canonical referent the participant can have no role
for link in participant_links:
role = link["role"] if link else None
link_id = int(link["xmi:id"]) if link else None
participants.append(
Participant(
text,
begin,
end,
element_id,
self.annotator_id,
self.title,
self,
role,
canonref,
link_id,
)
)
if participants:
self.participants = participants
# Create Filler objects
fillers = []
for fill in filler_xmidata:
text, begin, end, element_id = self.__extract_default_xmi(fill)
participant_links = list(
filter(
lambda fill_link: int(fill_link["target"]) == element_id,
fillerlink_xmidata,
)
)
# fill have 1 role each but if canonical referent the participant can have no role
for link in participant_links:
role = link["role"] if link else None
link_id = int(link["xmi:id"]) if link else None
fillers.append(
Filler(
text,
begin,
end,
element_id,
self.annotator_id,
self.title,
self,
role,
link_id,
)
)
if fillers:
self.fillers = fillers
# parse discontiguous trigger objects
discontiguous_triggers = []
for discont in discontiguous_xmidata:
text, begin, end, element_id = self.__extract_default_xmi(discont)
link = list(
filter(
lambda discont_link: int(discont_link["target"]) == element_id,
discontiguouslink_xmidata,
)
)
link_id = int(link[0]["xmi:id"]) if link else None
discontiguous_triggers.append(
DiscontiguousTrigger(
text,
begin,
end,
element_id,
self.annotator_id,
self.title,
self,
link_id,
)
)
if discontiguous_triggers:
self.discontiguous_triggers = discontiguous_triggers
# Create events objects
events = []
for event in events_xmidata:
text, begin, end, element_id = self.__extract_default_xmi(event)
event_type = event.get("a_Type", None)
event_subtype = event.get("b_Subtype", None)
event_fulltype = f"{event_type}.{event_subtype}"
# match participant by the link_id in the c_Participant feature
participant_link_id = event.get("c_Participant", None)
if participant_link_id: # some events have no participants
participant_link_id = [int(x) for x in participant_link_id.split()]
participants = list(
filter(
lambda p: p.link_id in participant_link_id, self.participants
)
)
else:
participants = None
# match fillers
filler_link_id = event.get("d_FILLER", None)
if filler_link_id: # some events have no fillers
filler_link_id = [int(x) for x in filler_link_id.split()]
fillers = list(
filter(lambda f: f.link_id in filler_link_id, self.fillers)
)
else:
fillers = None
# match discontiguous triggers
discont_link_id = event.get("f_Discontiguous", None)
if discont_link_id: # most events have no disconts
discont_link_id = [int(x) for x in discont_link_id.split()]
discontiguous_triggers = list(
filter(
lambda d: d.link_id in discont_link_id,
self.discontiguous_triggers,
)
)
else:
discontiguous_triggers = None
# parse factuality
othermodality = event.get("e_OtherModalityFactuality", None)
if othermodality == "false":
modality = "certain"
elif othermodality == "true":
modality = "other"
else:
modality = othermodality
negativepolarity = event.get("f_NegativePolarityFactuality", None)
if negativepolarity == "false":
polarity_negation = "positive"
elif negativepolarity == "true":
polarity_negation = "negative"
else:
polarity_negation = negativepolarity
if (
polarity_negation == "positive" and modality == "certain"
): # not the same as Liu
realis = "asserted"
else:
realis = "other"
# parse sentiment polarity
polarity_sentiment = event.get("i_Polarity", None)
if polarity_sentiment == "=Neutral":
polarity_sentiment = "neutral"
elif polarity_sentiment == "+Positive":
polarity_sentiment = "positive"
elif polarity_sentiment == "-Negative":
polarity_sentiment = "negative"
# set polarity when negated
if polarity_sentiment == "positive" and polarity_negation == "negative":
polarity_sentiment_scoped = "negative"
elif polarity_sentiment == "negative" and polarity_negation == "negative":
polarity_sentiment_scoped = "positive"
else: # for neutral or not negated
polarity_sentiment_scoped = polarity_sentiment
events.append(
Event(
text,
begin,
end,
element_id,
self.annotator_id,
self.title,
self,
event_type,
event_subtype,
event_fulltype,
discontiguous_triggers,
participants,
fillers,
polarity_negation,
modality,
realis,
polarity_sentiment,
polarity_sentiment_scoped,
)
)
if events:
self.events = events
# match coreferent events
self.coreferent_event_xmidata = coref_event_xmidata
for coref in coref_event_xmidata:
coref_from_id = int(coref["Governor"])
coref_to_id = int(coref["Dependent"])
from_event = next(
filter(lambda ev: ev.element_id == coref_from_id, self.events)
)
to_event = next(
filter(lambda ev: ev.element_id == coref_to_id, self.events)
)
if from_event.coreferents is not None:
if to_event not in from_event.coreferents:
from_event.coreferents.append(to_event)
else:
from_event.coreferents = [to_event]
if to_event.coreferents is not None:
if from_event not in to_event.coreferents:
to_event.coreferents.append(from_event)
else:
to_event.coreferents = [from_event]
# resolve coordinated events: coordinated events are in the exact same begin and end position
from operator import attrgetter
if self.events:
position_key = attrgetter("begin", "end")
evs = sorted(self.events, key=position_key)
coordinated = [
g
for g in [list(g) for k, g in groupby(evs, position_key)]
if len(g) > 1
]
for coord_event_group in coordinated:
for event in coord_event_group:
event.coordination = coord_event_group
# create Sentiment Expression objects
self.sentiment_expressions = []
self.targets = []
for sent in sentiment_xmidata:
(
sent_text,
sent_begin,
sent_end,
sent_element_id,
) = self.__extract_default_xmi(sent)
sent_polarity = sent.get("e_Polarity", None)
sent_uncertain = sent.get("c_Uncertain", None)
sent_negation = sent.get("d_Negated", None)
# parse sentiment polarity
if sent_polarity == "=Neutral":
sent_polarity = "neutral"
elif sent_polarity == "+Positive":
sent_polarity = "positive"
elif sent_polarity == "-Negative":
sent_polarity = "negative"
# flip polarity when negated
if sent_polarity == "positive" and sent_negation == "negative":
sent_polarity_scoped = "negative"
elif sent_polarity == "negative" and sent_negation == "negative":
sent_polarity_scoped = "positive"
else: # keep same for neutral or not negated
sent_polarity_scoped = sent_polarity
# create Target Links both Entity and Event targets
targets = []
# parse sentiment target links which link participants to sentiment as a target
targetentity_link_id = sent.get(
"a_Target"
).split() # split is needed because multiple target ids are space seperated
for tentlink_id in targetentity_link_id:
targetentity_link = next(
(
tlink
for tlink in targetentity_xmidata
if tlink["xmi:id"] == tentlink_id
),
None,
)
target_entity_id = targetentity_link["target"]
try:
target_participant = next(
p
for p in self.participants
if str(p.element_id) == target_entity_id
)
except StopIteration:
# target is not found in already parsed Event Participants, it is a new Participant annotation
part_xmi = next(
pxmi
for pxmi in participant_xmidata
if pxmi["xmi:id"] == target_entity_id
)
text, begin, end, element_id = self.__extract_default_xmi(part_xmi)
target_participant = Participant(
text,
begin,
end,
element_id,
self.annotator_id,
self.title,
self,