-
Notifications
You must be signed in to change notification settings - Fork 2
/
ccg_eval.py
539 lines (440 loc) · 15.9 KB
/
ccg_eval.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
import sys
import re
VERBOSE = False
RELATIONS = False
arg = 1
while arg < len(sys.argv):
if sys.argv[arg][0] != '-':
break
if sys.argv[arg] == '-v':
VERBOSE = True
elif sys.argv[arg] == '-r':
RELATIONS = True
arg += 1
(GOLD_LEXCATS, GOLD_DEPS, TEST, AUTO) = sys.argv[arg:]
eval_length = 0
COMMAND_LINE = ' '.join(sys.argv)
IGNORE = set(map(lambda x: tuple(x.split()), filter(lambda x: not x.startswith('#'), r"""
rule_id 7
rule_id 11
rule_id 12
rule_id 14
rule_id 15
rule_id 16
rule_id 17
rule_id 51
rule_id 52
rule_id 56
rule_id 91
rule_id 92
rule_id 95
rule_id 96
rule_id 98
conj 1 0
((S[to]{_}\NP{Z}<1>){_}/(S[b]{Y}<2>\NP{Z*}){Y}){_} 1 0
((S[to]{_}\NP{Z}<1>){_}/(S[b]{Y}<2>\NP{Z*}){Y}){_} 1 2
((S[to]{_}\NP{Z}<1>){_}/(S[b]{Y}<2>\NP{Z*}){Y}){_} 1 3
((S[to]{_}\NP{Z}<1>){_}/(S[b]{Y}<2>\NP{Z*}){Y}){_} 1 6
((S[to]{_}\NP{Z}<1>){_}/(S[b]{Y}<2>\NP{Z*}){Y}){_} 1 9
((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 1 6
((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_} 1 6
(((S[b]{_}\NP{Y}<1>){_}/PP{Z}<2>){_}/NP{W}<3>){_} 1 6
(S[X]{Y}/S[X]{Y}<1>){_} 1 13
(S[X]{Y}/S[X]{Y}<1>){_} 1 5
(S[X]{Y}/S[X]{Y}<1>){_} 1 55
((S[X]{Y}/S[X]{Y}){Z}\(S[X]{Y}/S[X]{Y}){Z}<1>){_} 2 97
((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_} 2 4
((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_} 2 93
((S[X]{Y}\NP{Z}){Y}\(S[X]{Y}<1>\NP{Z}){Y}){_} 2 8
((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_} 2 94
((S[X]{Y}\NP{Z}){Y}/(S[X]{Y}<1>\NP{Z}){Y}){_} 2 18
been ((S[pt]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 1 0
been ((S[pt]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 1 there 0
been ((S[pt]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 1 There 0
be ((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 1 there 0
be ((S[b]{_}\NP{Y}<1>){_}/NP{Z}<2>){_} 1 There 0
been ((S[pt]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 1 0
been ((S[pt]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 1 0
be ((S[b]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 1 0
have ((S[b]{_}\NP{Y}<1>){_}/(S[pt]{Z}<2>\NP{Y*}){Z}){_} 1 0
be ((S[b]{_}\NP{Y}<1>){_}/(S[adj]{Z}<2>\NP{Y*}){Z}){_} 1 0
be ((S[b]{_}\NP{Y}<1>){_}/(S[ng]{Z}<2>\NP{Y*}){Z}){_} 1 0
be ((S[b]{_}\NP{Y}<1>){_}/(S[pss]{Z}<2>\NP{Y*}){Z}){_} 1 0
going ((S[ng]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 1 0
have ((S[b]{_}\NP{Y}<1>){_}/(S[to]{Z}<2>\NP{Y*}){Z}){_} 1 0
Here (S[adj]{_}\NP{Y}<1>){_} 1 0
# this is a dependency Julia doesn't have but looks okay
from (((NP{Y}\NP{Y}<1>){_}/(NP{Z}\NP{Z}){W}<3>){_}/NP{V}<2>){_} 1 0
""".strip().split('\n'))))
deps_ignored = 0
def ignore(pred, cat, slot, arg, rule_id):
global deps_ignored
res = ('rule_id', rule_id) in IGNORE or \
(cat, slot, rule_id) in IGNORE or \
(pred, cat, slot, rule_id) in IGNORE or \
(pred, cat, slot, arg, rule_id) in IGNORE
deps_ignored += res
return res
MARKUP = re.compile(r'<[0-9]>|\{[A-Z_]\*?\}|\[X\]')
def strip_markup(cat):
cat = MARKUP.sub('', cat)
if cat[0] == '(':
return cat[1:-1]
else:
return cat
def next_gold_lexcats(lines, index):
lexcats = []
# line = file.readline()
# if not line:
# die("unexpected end of file reading gold standard lexical categories")
line = lines[index]
for (i, token) in enumerate(line.split()):
(word, pos, cat) = token.split('|')
lexcats.append((word, cat))
return lexcats, index + 1
def next_gold_deps(lines, index):
deps = set()
udeps = set()
line = lines[index]
if not line:
return (True, deps, udeps)
while line:
line = line.strip()
if not line:
break
(pred, cat, slot, arg) = line.split()
deps.add((pred, cat, slot, arg))
udeps.add((pred, arg))
index += 1
line = lines[index]
return (False, deps, udeps), index
def next_test(lines, index):
lexcats = []
deps = set()
udeps = set()
rule_ids = {}
# line = file.readline()
line = lines[index]
# if not line:
# die("unexpected end of file reading gold dependencies")
if line == '\n':
return (False, lexcats, deps, udeps, rule_ids), index + 1
while line:
line = line.strip()
if not line or line.startswith('<c>'):
break
fields = line.split()
(pred, cat, slot, arg, rule_id) = fields[:5]
pred_word = pred.rsplit('_')[0]
arg_word = arg.rsplit('_')[0]
if not ignore(pred_word, cat, slot, arg_word, rule_id):
cat = strip_markup(cat)
deps.add((pred, cat, slot, arg))
rule_ids[(pred, cat, slot, arg)] = rule_id
udeps.add((pred, arg))
index += 1
line = lines[index]
# if not line.startswith('<c>'):
# die("unexpected end of file reading test lexical categories")
for (i, token) in enumerate(line.split()[1:]):
(word, pos, cat) = token.split('|')
lexcats.append((word, cat))
# index += 1
# line = lines[index]
# if line != '\n':
# die("expected a blank line between each sentence")
if not rule_ids:
# No dependencies for this sentence - probably a conversion script error.
return (False, lexcats, deps, udeps, rule_ids), index + 1
return (True, lexcats, deps, udeps, rule_ids), index + 1
def get_valid_id():
id_list = []
with open(AUTO, 'r', encoding='utf8') as f:
lines = f.readlines()
for line in lines:
line = line.strip()
if line.startswith('ID='):
id_list.append(int(line[3:])-1)
return id_list
def get_auto():
pred_auto = []
with open(AUTO, 'r', encoding='utf8') as f:
lines = f.readlines()
for line in lines:
line = line.strip()
if line.startswith('ID='):
continue
leaves = re.findall("<L (.*?)>", line)
pairs = []
for leave in leaves:
items = leave.split()
assert len(items) == 5
ccg_categories = items[0]
modified_pos_tags = items[1]
original_pos_tags = items[2]
tokens = items[3]
predicate_arg_categories = items[4]
pairs.append((tokens, ccg_categories))
pred_auto.append(pairs)
return pred_auto
def score_deps(gold_deps, test_deps, rule_ids, verbose, relations,
correct_relations, incorrect_relations, missing_relations):
correct = gold_deps.intersection(test_deps)
if verbose:
for dep in correct:
print("correct: %s %s %s %s %s" % (dep + (rule_ids[dep],)))
if relations:
for dep in correct:
correct_relations[dep[1:3]] = correct_relations.setdefault(dep[1:3], 0) + 1
incorrect = test_deps.difference(gold_deps)
if verbose:
for dep in incorrect:
print("incorrect: %s %s %s %s %s" % (dep + (rule_ids[dep],)))
if relations:
for dep in incorrect:
incorrect_relations[dep[1:3]] = incorrect_relations.setdefault(dep[1:3], 0) + 1
missing = gold_deps.difference(test_deps)
if verbose:
for dep in missing:
print("missing: %s %s %s %s ?" % dep)
if relations:
for dep in missing:
missing_relations[dep[1:3]] = missing_relations.setdefault(dep[1:3], 0) + 1
# if verbose:
# print
return (len(correct), len(incorrect), len(missing))
def score_udeps(gold_deps, test_deps):
correct = gold_deps.intersection(test_deps)
incorrect = test_deps.difference(gold_deps)
missing = gold_deps.difference(test_deps)
return (len(correct), len(incorrect), len(missing))
def score_lexcats(gold_cats, test_cats):
if len(gold_cats) != len(test_cats):
return (0, 0)
# raise ValueError()
correct = 0
for (gold_w, gold_cat), (test_w, test_cat) in zip(gold_cats, test_cats):
if gold_w != test_w and not test_w.startswith('-'):
continue
# raise ValueError()
if gold_cat == test_cat:
correct += 1
return (len(gold_cats), correct)
# def read_preface(filename, file):
# preface = "# %s was generated using the following commands(s):\n" % filename
# line = file.readline()
# while line != "\n":
# if line.startswith("# this file"):
# line = file.readline()
# continue
# preface += line.replace('# ', '# ')
# line = file.readline()
# return preface
def read_preface(all_lines, index):
# preface = "# %s was generated using the following commands(s):\n" % filename
new_index = index
while all_lines[new_index] != "\n":
if all_lines[new_index].startswith("# this file"):
new_index += 1
continue
new_index += 1
return new_index + 1
preface = "# this file was generated by the following command(s):\n"
preface += "# %s\n" % COMMAND_LINE
with open(GOLD_LEXCATS, 'r', encoding='utf8') as f:
all_gold_lexcats_lines = f.readlines()
# gold_lexcats_index = 0
# gold_lexcats_index = read_preface(all_gold_lexcats_lines, gold_lexcats_index)
with open(GOLD_DEPS, 'r', encoding='utf8') as f:
all_gold_deps_lines = f.readlines()
# gold_deps_index = 0
# gold_deps_index = read_preface(all_gold_deps_lines, gold_deps_index)
with open(TEST, 'r', encoding='utf8') as f:
all_test_lines = f.readlines()
# gold_text_index = 0
# gold_text_index = read_preface(all_test_lines, gold_text_index)
nsentences = 0
parse_failures = 0
deps_sent_correct = 0
deps_correct = 0
deps_incorrect = 0
deps_missing = 0
udeps_sent_correct = 0
udeps_correct = 0
udeps_incorrect = 0
udeps_missing = 0
lexcats_sent_correct = 0
lexcats_total = 0
lexcats_correct = 0
relations_correct = {}
relations_incorrect = {}
relations_missing = {}
all_gold_deps_data = []
all_gold_lexcats_data = []
all_test_data = []
first_line_break = True
deps = set()
udeps = set()
for line in all_gold_deps_lines:
line = line.strip()
if line.startswith('#'):
continue
# if line == '' and first_line_break:
# first_line_break = False
# continue
if line == '':
if len(deps) > 0:
all_gold_deps_data.append((deps, udeps))
deps = set()
udeps = set()
continue
(pred, cat, slot, arg) = line.split()
deps.add((pred, cat, slot, arg))
udeps.add((pred, arg))
first_line_break = True
for line in all_gold_lexcats_lines:
line = line.strip()
if line == '' or line.startswith('#'):
continue
if line == '' and first_line_break:
first_line_break = False
continue
lexcats = []
for (i, token) in enumerate(line.split()):
(word, pos, cat) = token.split('|')
lexcats.append((word, cat))
all_gold_lexcats_data.append(lexcats)
first_line_break = True
lexcats = []
deps = set()
udeps = set()
rule_ids = {}
for line in all_test_lines:
line = line.strip()
if line.startswith('#'):
continue
if line.startswith('<c>n') and first_line_break:
first_line_break = False
continue
if line.startswith('<c>n'):
# for (i, token) in enumerate(line.split()[1:]):
# (word, pos, cat) = token.split('|')
# lexcats.append((word, cat))
if not rule_ids:
all_test_data.append((False, lexcats, deps, udeps, rule_ids))
else:
all_test_data.append((True, lexcats, deps, udeps, rule_ids))
lexcats = []
deps = set()
udeps = set()
rule_ids = {}
continue
else:
fields = line.split()
(pred, cat, slot, arg, rule_id) = fields[:5]
pred_word = pred.rsplit('_')[0]
arg_word = arg.rsplit('_')[0]
if not ignore(pred_word, cat, slot, arg_word, rule_id):
cat = strip_markup(cat)
deps.add((pred, cat, slot, arg))
rule_ids[(pred, cat, slot, arg)] = rule_id
udeps.add((pred, arg))
# if not line.startswith('<c>'):
# die("unexpected end of file reading test lexical categories")
pred_tags = get_auto()
valid_ids = get_valid_id()
assert len(valid_ids) == len(all_test_data)
assert len(all_gold_lexcats_data) == len(all_gold_deps_data)
all_gold_lexcats_data = [all_gold_lexcats_data[i] for i in valid_ids]
all_gold_deps_data = [all_gold_deps_data[i] for i in valid_ids]
assert len(all_test_data) == len(all_gold_lexcats_data)
for gold_lex, gold_dep, pred_dep, pred_tag in zip(all_gold_lexcats_data, all_gold_deps_data, all_test_data, pred_tags):
(gold_deps, gold_udeps) = gold_dep
(parsed, test_lexcats, test_deps, test_udeps, test_rule_ids) = pred_dep
if len(gold_lex) < eval_length:
continue
nsentences += 1
if not parsed:
parse_failures += 1
continue
(correct, incorrect, missing) = score_deps(gold_deps, test_deps, test_rule_ids, VERBOSE,
RELATIONS, relations_correct, relations_incorrect,
relations_missing)
deps_correct += correct
deps_incorrect += incorrect
deps_missing += missing
if incorrect == 0 and missing == 0:
deps_sent_correct += 1
(correct, incorrect, missing) = score_udeps(gold_udeps, test_udeps)
udeps_correct += correct
udeps_incorrect += incorrect
udeps_missing += missing
if incorrect == 0 and missing == 0:
udeps_sent_correct += 1
(total, correct) = score_lexcats(gold_lex, pred_tag)
lexcats_total += total
lexcats_correct += correct
if total == correct:
lexcats_sent_correct += 1
# print preface
print("note: all these statistics are over just those sentences")
print(" for which the parser returned an analysis, and ")
print(" dependency extraction script is successful \n")
def pct(val, total):
if val:
return 100.0 * val / total
else:
return 0.0
def print_acc(name, desc, correct, total):
acc = pct(correct, total)
print("%-6s %5.2f%% (%d of %d %s)" % (name + ':', acc, correct, total, desc))
def print_stats(name, correct, incorrect, missing):
test = correct + incorrect
prec = pct(correct, test)
print("%sp: %5.2f%% (%d of %d %s deps precision)" % (name[0], prec, correct, test, name))
gold = correct + missing
recall = pct(correct, gold)
print("%sr: %5.2f%% (%d of %d %s deps recall)" % (name[0], recall, correct, gold, name))
if prec and recall:
fscore = 2 * prec * recall / (prec + recall)
else:
fscore = 0.0
print("%sf: %5.2f%% (%s deps f-score)" % (name[0], fscore, name))
def print_rel_stats(relation, correct, incorrect, missing):
relation = "%s %s" % relation
test = correct + incorrect
prec = pct(correct, test)
gold = correct + missing
recall = pct(correct, gold)
if prec and recall:
fscore = 2 * prec * recall / (prec + recall)
else:
fscore = 0.0
print("%-50s: %6.2f%% %6.2f%% %6.2f%% %6d %6d" % (relation, prec, recall, fscore, test, gold))
# if RELATIONS:
# relations = relations_correct.copy()
# for (r, freq) in relations_missing:
# relations[r] = relations.get(r, 0) + int(freq)
#
# relations = map(lambda x: (x[1], x[0]), relations.items())
# relations = sorted(list(relations), key=lambda kv: kv[0])
#
# for (freq, r) in relations:
# print_rel_stats(r, relations_correct.get(r, 0), relations_incorrect.get(r, 0),
# relations_missing.get(r, 0))
# print('')
nparsed = nsentences - parse_failures
print_acc("cover", "sentences evaluated - this includes dependency extraction script errors and parse failures",
nparsed, nsentences)
print('')
print_acc("cats", "tokens correct", lexcats_correct, lexcats_total)
print_acc("csent", "sentences correct", lexcats_sent_correct, nparsed)
print('')
print_stats("labelled", deps_correct, deps_incorrect, deps_missing)
print_acc("lsent", "labelled deps sentences correct", deps_sent_correct, nparsed)
print('')
print_stats("unlabelled", udeps_correct, udeps_incorrect, udeps_missing)
print_acc("usent", "unlabelled deps sentences correct", udeps_sent_correct, nparsed)
print('')
print_acc("skip", "ignored deps (to ensure compatibility with CCGbank)", deps_ignored,
deps_correct + deps_incorrect + deps_ignored)