-
Notifications
You must be signed in to change notification settings - Fork 2
/
quotedspeechpredictor.py
645 lines (601 loc) · 28.6 KB
/
quotedspeechpredictor.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
import vozbase
import voz
import quotedspeechhelper
import styhelper
import settings
import logging
import entitymanager
import verbmanager
import sys
logger = logging.getLogger(__name__)
import random,collections
import copy
import util
import utterancegenerator
from quotedbase import *
import itertools
def tokenize_document(doc, verbose=False):
assert isinstance(doc,voz.Document)
quotes = [] #type: list[Quote]
mentions = [] #type: list[entitymanager.Mention]
verbs = [] #type: list[verbmanager.Verb]
# parse
in_quotes_offset = None
output = []
input = doc.get_all_tokens(pair_container=True)
ignore_tokens = set()
while input:
sentence, token = input.pop(0)
assert isinstance(sentence,voz.Sentence)
assert isinstance(token,voz.Token)
if in_quotes_offset is None and token.text[-1] in ['.',':']:
output.append(Punctuation(token.text[-1]))
elif in_quotes_offset is None and token.text!='``':
# decide what to do with tokens outside of quotes
if not token in ignore_tokens:
mentions_ = filter(None,doc.get_mentions_by_token_id(token.id))
mentions_ = util.flatten([i.get_all_children_recursively() for i in mentions_])
for mention in set(mentions_):
if mention and mention.is_independent:
if mention.get_most_likely_symbol():
output.append(mention)
mentions.append(mention)
ignore_tokens.update(mention.tokens)
verb = doc.get_verb_by_token_id(token.id)
if verb:
# TODO whitelist only expression verbs?
verbs.append(verb)
output.append(verb)
elif in_quotes_offset is None and token.text=='``':
in_quotes_offset = token.offset
elif in_quotes_offset is not None and token.text=="''":
quote = Quote(in_quotes_offset,token.offset+token.len, doc)
quote.endp = token.get_previous().text
quotes.append(quote)
output.append(quote)
if sentence.annotations and sentence.annotations.speech_data:
quote.annotations = sentence.annotations.get_speech_annotations_for_quote(quote)
else:
logger.warning("NO ANNOTATION DATA FOR SENTENCE")
in_quotes_offset = None
return output,quotes,mentions,verbs
def predict_quoted_speech_informed(input_tuple, overwrite_mode_default=True, rule_name='baseline'):
output, quotes,mentions,verbs = input_tuple
symbol_to_mention = collections.defaultdict(list)
for mention in mentions:
symbol = mention.get_most_likely_symbol()
if symbol:
symbol_to_mention[symbol].append(mention)
s = sorted([(len(v),k,v) for k,v in symbol_to_mention.items()],reverse=True)
a = s[0][2][0]
b = s[1][2][0]
for quote in quotes:
if overwrite_mode_default:
quote.set_speaker_mention(a,rule_name)
quote.set_listener_mention(b,rule_name)
else:
if not quote.speaker_mention:
quote.speaker_mention = a
if not quote.listener_mention:
quote.listener_mention = b
return output, quotes,mentions,verbs
def weighted_assignment(input_tuple,weights,top_weight_mechanism=0):
output, quotes, mentions, verbs = input_tuple
def top(items,slot):
return sorted(items,key=lambda i:(weights[i[0]][slot] if i[0] in weights else 0.0))[-1][1]
def weight(items,slot):
cumm_weights = {}
total = 0.0
for rule_name,mention in items:
cumm_weights[mention] = cumm_weights.get(mention,0.0) + weights[rule_name][slot]
return sorted(cumm_weights.items(),key=lambda i:i[1])[-1][0]
def bayes(items,slot):
mentions = [i[1] for i in items]
cumm_weights = dict([(i,0.0) for i in mentions])
for rule_name,mention in items:
if not rule_name in weights: continue
cumm_weights[mention] += weights[rule_name][slot]
for other in mentions:
if other==mention: continue
cumm_weights[other] += (1.0-weights[rule_name][slot])/(len(mentions)-1)
return sorted(cumm_weights.items(), key=lambda i: i[1])[-1][0]
def counter(items,slot):
return collections.Counter([i[1] for i in items]).most_common(1)[0][0]
if top_weight_mechanism==0:
selector = top
elif top_weight_mechanism==1:
selector = weight
elif top_weight_mechanism==2:
selector = bayes
elif top_weight_mechanism==3:
selector = counter
for quote in quotes:
quote.speaker_mention = None
quote.listener_mention = None
if quote.speaker_mentions:
quote.speaker_mention = selector(quote.speaker_mentions.items(),0)
if quote.listener_mentions:
quote.listener_mention = selector(quote.listener_mentions.items(),1)
def eval_quoted_speech(input_tuple,rule=None, laplace = 0, verbose=False):
output, quotes,mentions,verbs = input_tuple
s_ms = laplace # matching
s_ml = laplace
s_ts = laplace * 2 # total checked
s_tl = laplace * 2
s_cs = 0 # correct annotations
s_cl = 0
s_os = 0 # only rule
s_ol = 0
s_as = 0 # agreement
s_al = 0
s_ns = 0 # conflict
s_nl = 0
for quote in quotes:
if not rule:
pass
'''if quote.speaker_mentions:
quote.speaker_mention = quote.speaker_mentions.values()[0]
if quote.listener_mentions:
quote.listener_mention = quote.listener_mentions.values()[0]'''
else:
quote.speaker_mention = None
quote.listener_mention = None
if quote.speaker_mentions and rule in quote.speaker_mentions:
quote.speaker_mention = quote.speaker_mentions[rule]
if len(quote.speaker_mentions)==1:
s_os +=1
else:
s_as = quote.speaker_mentions.values().count(quote.speaker_mention) - 1
s_ns = len(quote.speaker_mentions.values()) - s_as - 1
if quote.listener_mentions and rule in quote.listener_mentions:
quote.listener_mention = quote.listener_mentions[rule]
if len(quote.speaker_mentions) == 1:
s_ol += 1
else:
s_al = quote.speaker_mentions.values().count(quote.speaker_mention) - 1
s_nl = len(quote.speaker_mentions.values()) - s_al - 1
if quote.annotations.speaker_annotation and ',' not in quote.annotations.speaker_annotation:
s_cs +=1
if quote.speaker_mention:
s_ts +=1
if quote.speaker_mention.get_most_likely_symbol() == quote.annotations.speaker_annotation:
s_ms += 1
else:
logger.info('WRONG SPEAKER ASSIGNMENT %d, %s' % (quote.doc.id, str(quote)))
else:
logger.info('WRONG ANNOTATION IN SPEAKER QUOTE %d, %s' % (quote.doc.id, str(quote)))
if quote.annotations.listener_annotation and ',' not in quote.annotations.listener_annotation:
s_cl += 1
if quote.listener_mention:
s_tl +=1
if quote.listener_mention.get_most_likely_symbol() == quote.annotations.listener_annotation:
s_ml += 1
else:
logger.info('WRONG LISTENER ASSIGNMENT %d, %s' % (quote.doc.id, str(quote)))
else:
pass
logger.info('WRONG ANNOTATION IN LISTENER QUOTE %d, %s' % (quote.doc.id, str(quote)))
if verbose:
print quote.speaker_mention.get_most_likely_symbol() if quote.speaker_mention else '?',quote.annotations.speaker_annotation,'>', quote.listener_mention.get_most_likely_symbol() if quote.listener_mention else '?', quote.annotations.listener_annotation
return s_ms, s_ml, s_ts, s_tl, s_cs, s_cl, len(quotes), s_os, s_as, s_ns, s_ol, s_al, s_nl
def print_eval_quoted_speech(input_tuple,verbose=False):
s_ms, s_ml, s_ts, s_tl, s_cs, s_cl, len_quotes, s_os, s_as, s_ns, s_ol, s_al, s_nl = input_tuple
print 'ACCURACY (SPEAKER,LISTENER,TOTAL):',1.0*s_ms/s_ts if s_ts else 0.0,1.0*s_ml/s_tl if s_tl else 0.0,0.5*(s_ms+s_ml)/s_ts if s_ts else 0.0
print 'COVERAGE (SPEAKER,LISTENER,TOTAL):', 1.0 * s_ts / s_cs, 1.0 * s_tl/s_cl, 1.0 * (s_ts + s_tl) / (s_cs+ s_cl), 'SKIPPED', 1.0*(len_quotes - s_cs)/len_quotes, 1.0*(len_quotes - s_cl)/len_quotes
def compute_eval_quoted_speech(input_tuple,verbose=False):
s_ms, s_ml, s_ts, s_tl, s_cs, s_cl, len_quotes, s_os, s_as, s_ns, s_ol, s_al, s_nl = input_tuple
return 'ACCURACY',1.0*s_ms/s_ts if s_ts else 0.0,1.0*s_ml/s_tl if s_tl else 0.0,0.5*(s_ms+s_ml)/s_ts if s_ts else 0.0,'COVERAGE',1.0 * s_ts / s_cs, 1.0 * s_tl/s_cl, 1.0 * (s_ts + s_tl) / (s_cs+ s_cl),'SKIPPED', 1.0*(len_quotes - s_cs)/len_quotes, 1.0*(len_quotes - s_cl)/len_quotes, 'CONFLICTS',s_os, s_as, s_ns, s_ol, s_al, s_nl, 'MATCHES', s_ms, s_ml
class QuotedSpeechPredictorRule(object):
def __init__(self, pattern, actions, rule_type=None):
self.pattern = pattern
self.actions = actions
self.rule_type = rule_type
@classmethod
def from_string(cls, s, t=None):
if t:
s = s.strip(t)
logger.info("Loading rule: %s" % s )
pattern,actions = s.split('>')
pattern = pattern.strip().split()
actions = actions.strip().split()
actions = [i.replace('=','.').split('.') for i in actions]
return cls(pattern, actions, t)
@classmethod
def format_action(cls, action):
if len(action) == 3:
return '%s.%s=%s' % tuple(action)
elif len(action) == 4:
return '%s.%s=%s.%s' % tuple(action)
else:
return 'ERR'
def __str__(self):
return (self.rule_type or '') + ' '.join(self.pattern) + ' > ' + ' '.join([QuotedSpeechPredictorRule.format_action(i) for i in self.actions])
def __eq__(self,other):
return hash(self)==hash(other) and str(self)==str(other)
def __ne__(self,other):
return not self == other
def __hash__(self):
return hash(str(self))
class QuotedSpeechMatcher(object):
property_setters = {'s':'set_speaker_mention','l':'set_listener_mention','m':'add_matched_rules'}
property_getters = {'s': 'get_speaker_mention', 'l': 'get_listener_mention'}
def __init__(self, rule, tokens = None):
self.rule = rule
self.tokens = tokens or []
def clone(self):
return QuotedSpeechMatcher(self.rule,self.tokens)
@classmethod
def match(cls, token, target):
t = target[0]
if t=='Q' and isinstance(token, Quote): return True
if t=='P' and isinstance(token, entitymanager.Mention): return True
if t=='E' and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_EXPRESS: return True
if t=='S' and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_SAY: return True
if t=='A' and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_ASK: return True
if t=='R' and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_REPLY: return True
if t=='T' and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_THINK: return True
if t=='V' and isinstance(token, verbmanager.Verb) : return True
if t==':' and isinstance(token, Punctuation) and token.endp == ':': return True
if t=='.' and isinstance(token, Punctuation) and token.endp == '.': return True
if t=='q' and target.startswith('q~') and isinstance(token, Quote) and token.endp[-1] not in ['.','!','?']: return True
if t=='q' and target.startswith('q!') and isinstance(token, Quote) and token.endp[-1] in ['.','!','?']: return True
return False
if target.startswith('Q') and isinstance(token, Quote): return True
if target.startswith('P') and isinstance(token, entitymanager.Mention): return True
if target.startswith('E') and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_EXPRESS: return True
if target.startswith('S') and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_SAY: return True
if target.startswith('A') and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_ASK: return True
if target.startswith('R') and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_REPLY: return True
if target.startswith('T') and isinstance(token, verbmanager.Verb) and token.token.lemma in VERBS_THINK: return True
if target.startswith('V') and isinstance(token, verbmanager.Verb) : return True
if target.startswith(':') and isinstance(token, Punctuation) and token.endp == ':': return True
if target.startswith('.') and isinstance(token, Punctuation) and token.endp == '.': return True
# [('said', 202), ('asked', 46), ('answered', 22), ('cried', 21), ('thought', 14), ('replied', 10), ('called', 9), ('saying', 7), ('told', 5), ('begged', 4), ('sang', 4), ('yelled', 4), ('whispered', 3), ('repeated', 3), ('croaked', 1), ('urge', 1), ('ran', 1), ('issued', 1), ('winked', 1), ('related', 1), ('ordered', 1), ('hissing', 1), ('muttered', 1), ('coaxed', 1), ('asks', 1), ('implore', 1), ('began', 1), ('asking', 1), ('questioned', 1), ('crying', 1), ('roared', 1), ('repeating', 1), ('Saying', 1), ('cuckooed', 1), ('warned', 1), ('groaned', 1), ('beg', 1), ('exclaimed', 1), ('went', 1)]
@classmethod
def match_first(cls, rule, token):
target = rule.pattern[0]
if target[0]=='?': return True
return cls.match(token, target)
def matched(self):
return len(self.tokens) >= len(self.rule.pattern)
def ingest(self, token):
while True:
target = self.rule.pattern[len(self.tokens)]
if target[0]=='?':
target = target[1:]
optional = True
else:
optional = False
if self.match(token, target):
self.tokens.append(token)
return self
else:
if optional:
self.tokens.append(EmptyTokenizedToken(target))
# check this is not the last target
if self.matched():
return self
else:
return None
'''def ingest_old_that_returns_lists(self, token):
ret = []
current_matcher = self
while True:
target = current_matcher.rule.pattern[len(current_matcher.tokens)]
if not target.startswith('?'): break
current_matcher = current_matcher.clone()
current_matcher.tokens.append(EmptyTokenizedToken(target))
ret.append(current_matcher)
if self.match(token, target):
self.tokens.append(token)
ret.append(self)
else:
ret.append(None)
return ret
def ingest_old_w_o_support_for_optional(self, token):
target = self.rule.pattern[len(self.tokens)]
if self.match(token, target):
self.tokens.append(token)
return self
else:
return None'''
def try_apply(self):
if self.matched():
self.apply()
return None
else:
return self
def apply(self, verbose=False):
if verbose:
print 'APPLYING RULE',self.rule
if verbose:
for token in self.tokens:
if isinstance(token,Quote):
print token
named = dict(zip(self.rule.pattern,self.tokens))
named.update(dict(zip([i.replace('q~','Q').replace('q!','Q') for i in self.rule.pattern], self.tokens)))
for action in self.rule.actions:
#print self.rule
if len(action)==3:
# assignment using an object
quote_name, p_set, target_name = action
val = named[target_name]
elif len(action)==4:
# assignment using an object and a getter
quote_name, p_set, target_name, p_get = action
val = getattr(named[target_name], self.property_getters[p_get])()
getattr(named[quote_name], self.property_setters[p_set])(val, str(self.rule))
def predict_quoted_speech(input_tuple, rules):
output, quotes,mentions,verbs = input_tuple
apply_matchers(output,rules)
return output, quotes, mentions, verbs
def apply_matchers(tokenized_string,rules):
matchers = []
for token in tokenized_string:
for rule in rules:
if QuotedSpeechMatcher.match_first(rule, token):
matchers.append(QuotedSpeechMatcher(rule))
matchers = [matcher.ingest(token) for matcher in matchers]
#matchers = util.flatten(matchers) # necessary to support ND-FSA
matchers = filter(None, matchers) # kill dead matchers
matchers = [matcher.try_apply() for matcher in matchers]
matchers = filter(None, matchers) # kill matchers that applied successfully
def extract_rules_window(output, quote, token_start, token_end):
rules = []
tokens = output[token_start:token_end]
# check if there are valid mentions to construct the rule
mentions = [i for i in tokens if isinstance(i,entitymanager.Mention)]
speakers = [i for i in mentions if i.get_most_likely_symbol()==quote.annotations.speaker_annotation and quote.annotations.speaker_annotation]
listeners = [i for i in mentions if i.get_most_likely_symbol()==quote.annotations.listener_annotation and quote.annotations.listener_annotation]
quotes = [i for i in tokens if isinstance(i,Quote) and i!=quote]
# compute rules for the given quote
pattern = ['%s%d' % (token_to_string(token),i) for i,token in enumerate(tokens)]
# TODO do it for all the quotes in the token list?
token_to_string_dict = dict(zip(tokens,pattern))
if quote not in token_to_string_dict:
pass
target = token_to_string_dict[quote]
if speakers or listeners:
actions = []
for p in speakers:
actions.append((target,'s',token_to_string_dict[p]))
for p in listeners:
actions.append((target,'l',token_to_string_dict[p]))
rules.append(QuotedSpeechPredictorRule(pattern,actions))
if quotes:
actions = []
for q in quotes:
if quote.annotations.speaker_annotation==q.annotations.speaker_annotation:
actions.append((target,'s',token_to_string_dict[q],'s'))
if quote.annotations.speaker_annotation==q.annotations.listener_annotation:
actions.append((target,'s',token_to_string_dict[q],'l'))
if quote.annotations.listener_annotation==q.annotations.speaker_annotation:
actions.append((target,'l',token_to_string_dict[q],'s'))
if quote.annotations.listener_annotation==q.annotations.listener_annotation:
actions.append((target,'l',token_to_string_dict[q],'l'))
rules.append(QuotedSpeechPredictorRule(pattern, actions, FOLLOWUP_RULE))
return rules
def extract_rules(input, windows_size=8):
rules = []
missing = []
output, quotes, mentions, verbs = input
for token_i, token in enumerate(output):
if isinstance(token,Quote):
rules_ = []
for before in range(0,windows_size):
for after in range(1,windows_size):
start = token_i - before
end = token_i + after
if start<0 or end>len(output): continue
if not token.annotations: continue
rules_ += extract_rules_window(output,token,start, end)
if not rules_:
missing += [token]
rules += rules_
return rules, missing
def clean_assignments(data_set):
for output_tuple in data_set.values():
output, quotes, mentions, verbs = output_tuple
for quote in quotes:
quote.speaker_mention = None
quote.speaker_mentions = {}
quote.listener_mention = None
quote.listener_mentions = {}
def generalize_rules(rules):
def generalize_options(i, actions):
r = []
if i[0] in ['S', 'A', 'R', 'T', 'E', 'q']:
pass
#r.append(generalize_token(i))
if i not in [j[0] for j in actions] and i not in [j[2] for j in actions]:
r.append('?' + i)
if not r:
r = [i]
return r
def generalize_token(i):
return RULE_LANGUAGE.get(i[0:-1], i[0:-1]) + i[-1]
def generalize_action(j):
j = list(j)
j[0] = generalize_token(j[0])
return j
rules2 = []
for rule in rules:
patterns = itertools.product(*[generalize_options(i, rule.actions) for i in rule.pattern])
for pattern2 in patterns:
actions2 = [j if (j[0] in pattern2) else generalize_action(j) for j in rule.actions]
actions2 = [j for j in actions2 if j[2] in pattern2 and j[0] in pattern2]
if not actions2: continue
rules2.append(QuotedSpeechPredictorRule(pattern2, actions2, rule.rule_type))
return rules2
def main_minilang_parser_test():
tokenized = [Punctuation('.'), Punctuation(':'), Punctuation('.'), Punctuation(':')]
rules = [
':1 . :2 > :1.m=:2',
':1 ?. :2 > :1.m=:2',
'?. ?. :1 ?. ?. :2 > :1.m=:2'
]
rules = [QuotedSpeechPredictorRule.from_string(i) for i in rules]
apply_matchers(tokenized,rules)
for i in tokenized:
print i.endp, len(i.matched_rules), i.matched_rules
def minilang_permutations(lst):
return '('+'|'.join([' '.join(list(i)) for i in itertools.permutations(lst)])+')'
def load_rules_manual():
# example rules
# Q S P > Q.s=P
# when Quote Said Person is found, assign Person to Quote.speaker
# Q1 S P1 Q2 R P2 > Q1.s=P1 Q1.l=P2 Q2.s=P2 Q2.l=P1:
# when Quote1 Said Person1 Quote2 Replied Person2 is found, assign Person1 to Quote1.speaker, Person2 to Quote1.listener...
rules = [
'Q S P > Q.s=P',
'Q P S > Q.s=P',
'P S Q > Q.s=P',
'S P Q > Q.s=P',
'S Q P > Q.s=P',
'P Q S > Q.s=P',
'Q1 S P1 Q2 R P2 > Q1.s=P1 Q1.l=P2 Q2.s=P2 Q2.l=P1',
'Q E P > Q.s=P',
'Q P E > Q.s=P',
'P E Q > Q.s=P',
'E P Q > Q.s=P',
'E Q P > Q.s=P',
'P Q E > Q.s=P',
'Q T ?: P > Q.s=P Q.l=P',
'Q P T ?. > Q.s=P Q.l=P',
'P T ?: Q > Q.s=P Q.l=P',
'T P ?: Q > Q.s=P Q.l=P',
'T Q P ?. > Q.s=P Q.l=P',
'P Q T ?. > Q.s=P Q.l=P',
'Q1 E P1 Q2 R P2 > Q1.s=P1 Q1.l=P2 Q2.s=P2 Q2.l=P1',
'Q S P > Q.s=P',
'q~ S P > q~.s=P',
'q~ S P q! > q~.s=P q!.s=P',
#'q~1 S P q~2 q!> q~1.s=P q~2.s=P q!.s=P',
'P S q! > q!.s=P',
'P R q! > q!.s=P',
#'P1 S q!1 P2 R q!2 > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1', doesn't work with punctuation
'P1 S q!1 P2 R q!2 > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1',
'q!1 q!2 P A > q!1.l=P q!2.s=P',
'q!1 q~ S P q!2 > Q.s=P Q1.s=P Q2.s=P',
'P E : Q > Q.s=P',
'E P : Q > Q.s=P',
'P S : Q > Q.s=P',
'S P : Q > Q.s=P',
'P A : Q > Q.s=P',
#'A P : Q > Q.s=P',
'P1 E P2 : Q > Q.s=P1 Q.l=P2',
'P2 E P1 : Q > Q.s=P1 Q.l=P2',
'P1 S P2 : Q > Q.s=P1 Q.l=P2',
#'P2 S P1 : Q > Q.s=P1 Q.l=P2',
'P1 A P2 : Q > Q.s=P1 Q.l=P2',
'P2 A P1 : Q > Q.s=P1 Q.l=P2',
'P1 E ?: q!1 P2 E ?: q!2 > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1',
'P1 E ?: q!1 P2 R ?: q!2 > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1',
'P1 A ?: q!1 P2 R ?: q!2 > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1',
'P1 E ?: q!1 q!2 P2 E > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1',
'P1 A ?: q!1 q!2 P2 R > q!1.s=P1 q!1.l=P2 q!2.s=P2 q!2.l=P1',
'P1 E P2 ?: Q > Q.s=P1 Q.l=P2',
# new:
# 'P1 V V P2 ?: Q > Q.s=P1 Q.l=P2', #Then the dragon began to implore Nikita:
]
rules_ = [QuotedSpeechPredictorRule.from_string(i) for i in rules]
rules = [
'_Q1 Q2 > Q1.l=Q2.s Q2.l=Q1.s',
'_Q2 Q1 > Q1.l=Q2.s Q2.l=Q1.s',
'_Q1 ?V ?P ?V ?P Q2 > Q1.l=Q2.s Q2.l=Q1.s',
'_Q2 ?V ?P ?V ?P Q1 > Q1.l=Q2.s Q2.l=Q1.s',
'_q~1 ?V ?P ?V ?P q!2 > Q2.s=Q1.s Q2.l=Q1.l',
'_q~2 ?V ?P ?V ?P q!1 > Q2.s=Q1.s Q2.l=Q1.l',
]
return rules_ + [QuotedSpeechPredictorRule.from_string(i,FOLLOWUP_RULE) for i in rules]
def load_rules_from_permutation_minilang(use_qsa_subset=False):
d = {'Q': ['Q','q~','q!'],
'E': ['E','A','R','T','S'],
}
if not use_qsa_subset:
g = [
minilang_permutations(['{Q}', 'P', '{E}']) + ' ?. > Q.s=P',
minilang_permutations(['{Q}1', 'P1', '{E}']) + ' ?. ' + minilang_permutations(['{Q}2', 'P2', '{E}']) + ' > Q1.s=P1 Q2.s=P2',
minilang_permutations(['{Q}1', 'P1', '{E}']) + ' ?. ' + minilang_permutations(['{Q}2', 'P2', '{E}']) + ' > Q1.s=P1 Q2.s=P2 Q1.l=P2 Q2.l=P1',
'(P {E}|{E} P) ?: {Q} ?. > Q.s=P',
'(P1 {E}|{E} P1) ?: {Q}1 {Q}2 (P2 {E}|{E} P2) ?. > Q1.s=P1 Q2.s=P2 Q1.l=P2 Q2.l=P1',
'(P1 {E}|{E} P1) ?: {Q}1 (P2 {E}|{E} P2) ?: {Q}2 ?. > Q1.s=P1 Q2.s=P2 Q1.l=P2 Q2.l=P1',
'{Q}1 {Q}2 (P {E}|{E} P) ?. > Q2.s=P Q1.l=P',
'{Q}1 {Q}2 (P {E}|{E} P) ?. {Q}3 > Q2.s=P Q1.l=P Q3.l=P',
'P1 {E} P2 ?: Q > Q.s=P1 Q.l=P2',
'q~1 (P {E}|{E} P) q!2 > Q1.s=P Q2.s=P',
'P ?P2 ?V . {Q} > Q.s=P',
'P ?V . {Q} > Q.s=P',
'P1 ?V P2 ?V . {Q} > Q.s=P1 Q.l.P2',
'P2 ?V P1 ?V . {Q} > Q.s=P1 Q.l.P2',
'P1 ?V P2 ?V . {Q}1 {Q}2 > Q1.s=P1 Q1.l.P2 Q2.s=P2 Q2.l.P1',
'P2 ?V P1 ?V . {Q}1 {Q}2 > Q1.s=P1 Q1.l.P2 Q2.s=P2 Q2.l.P1',
'P ?E ?V ?V ?V ?: {Q} > Q.s=P',
'P1 ?V (V|E) ?V P2 ?: {Q} > Q.s=P1 Q.l=P2',
'P2 ?V (V|E) ?V P1 ?: {Q} > Q.s=P1 Q.l=P2',
]
else:
g = [
minilang_permutations(['{Q}', 'P', '{E}']) + ' ?V ?. ?: > Q.s=P',
minilang_permutations(['{Q}1', 'P1', '{E}']) + ' ?V ?. ?: ' + minilang_permutations(
['{Q}2', 'P2', '{E}']) + ' > Q1.s=P1 Q2.s=P2',
minilang_permutations(['{Q}1', 'P1', '{E}']) + ' ?V ?. ?: ' + minilang_permutations(
['{Q}2', 'P2', '{E}']) + ' > Q1.s=P1 Q2.s=P2 Q1.l=P2 Q2.l=P1',
]
rules = utterancegenerator.UtteranceGenerator().generate(g, d, verbose=False)
return [QuotedSpeechPredictorRule.from_string(i) for i in rules]
def main_gene_minilang():
d = {'Q': ['Q','q~','q!'],
'E': ['E','A','R','T','S'],
}
g = [
minilang_permutations(['{Q}', 'P', '{E}']) + ' (?.|)',
minilang_permutations(['{Q}', 'P', '{E}']) + ' ' + minilang_permutations(['{Q}', 'P', '{E}']),
'(P {E}|{E} P) (?:|:|) {Q} (?.|)',
'(P {E}|{E} P) (?:|:|) {Q} {Q} (P {E}|{E} P) (?.|)',
'(P {E}|{E} P) (?:|:|) {Q} (P {E}|{E} P) (?:|:|) {Q} (?.|)',
'{Q} {Q} (P {E}|{E} P) (?.|)',
'{Q} {Q} (P {E}|{E} P) (?.|) {Q}',
'P {E} P (?:|:|) Q',
'q~ (P {E}|{E} P) q!',
]
results = utterancegenerator.UtteranceGenerator().generate(g, d, verbose=False)
results = [i.strip() for i in results]
print 'generated rules',len(results),results
def main_print_stats():
len_quotes = 0
len_sentences = 0
len_verbs = 0
len_mentions = 0
len_pp = 0
len_pn = 0
len_tokens = 0
len_tokens_in_quotes = 0
for story_file in settings.STY_FILES:
print story_file
doc = styhelper.create_document_from_sty_file(settings.STY_FILE_PATH+story_file)
#styhelper.fix_sty_annotations(doc)
quotedspeechhelper.annotate_sentences(doc, settings.STORY_ALL_SENTENCES, single_sentences_file_story_id = doc.id)
output_tuple = tokenize_document(doc)
output, quotes, mentions, verbs = output_tuple
print tokenized_string_to_string(output, 1)
len_quotes += len(quotes)
len_verbs += len(verbs)
len_mentions += len(mentions)
len_sentences += len(doc.sentences)
len_pp += len([i for i in mentions if [j for j in i.tokens if j.pos == 'PRP']])
len_pn += len([i for i in mentions if [j for j in i.tokens if j.pos == 'NNP']])
len_tokens += len(doc.get_text())
len_tokens_in_quotes += sum([q.offset_end-q.offset for q in quotes])
print 'TOTAL NUM QUOTES\t', len_quotes
print 'TOTAL NUM SENT\t', len_sentences
print 'TOTAL NUM VERBS\t', len_verbs
print 'TOTAL NUM MENT\t', len_mentions
print 'TOTAL NUM PP\t', len_pp
print 'TOTAL NUM PN\t', len_pn
print 'TOTAL NUM chars\t', len_tokens
print 'TOTAL NUM chars in quotes\t', len_tokens_in_quotes
if __name__=='__main__':
main_minilang_parser_test()
main_gene_minilang()
main_print_stats()