forked from WING-NUS/sequicity
-
Notifications
You must be signed in to change notification settings - Fork 15
/
Copy pathmetric.py
executable file
·664 lines (580 loc) · 25.7 KB
/
metric.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
import csv
from collections import Counter
from nltk.util import ngrams
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
from nltk.stem import WordNetLemmatizer
import math, re, argparse
import json
import functools
import pickle
from reader import clean_replace
from config import global_config as cfg
import logging
import pdb
en_sws = set(stopwords.words())
wn = WordNetLemmatizer()
order_to_number = {
'first': 1, 'one': 1, 'seco': 2, 'two': 2, 'third': 3, 'three': 3, 'four': 4, 'forth': 4, 'five': 5, 'fifth': 5,
'six': 6, 'seven': 7, 'eight': 8, 'nin': 9, 'ten': 10, 'eleven': 11, 'twelve': 12
}
def similar(a,b):
return a == b or a in b or b in a or a.split()[0] == b.split()[0] or a.split()[-1] == b.split()[-1]
#return a == b or b.endswith(a) or a.endswith(b)
def setsub(a,b):
junks_a = []
useless_constraint = ['temperature','week','est ','quick','reminder','near']
for i in a:
flg = False
for j in b:
if similar(i,j):
flg = True
if not flg:
junks_a.append(i)
for junk in junks_a:
flg = False
for item in useless_constraint:
if item in junk:
flg = True
if not flg:
return False
return True
def setsim(a,b):
a,b = set(a),set(b)
return setsub(a,b) and setsub(b,a)
class BLEUScorer(object):
## BLEU score calculator via GentScorer interface
## it calculates the BLEU-4 by taking the entire corpus in
## Calulate based multiple candidates against multiple references
def __init__(self):
pass
def score(self, parallel_corpus):
# containers
count = [0, 0, 0, 0]
clip_count = [0, 0, 0, 0]
r = 0
c = 0
weights = [0.25, 0.25, 0.25, 0.25]
# accumulate ngram statistics
for hyps, refs in parallel_corpus:
hyps = [hyp.split() for hyp in hyps]
refs = [ref.split() for ref in refs]
for hyp in hyps:
for i in range(4):
# accumulate ngram counts
hypcnts = Counter(ngrams(hyp, i + 1))
cnt = sum(hypcnts.values())
count[i] += cnt
# compute clipped counts
max_counts = {}
for ref in refs:
refcnts = Counter(ngrams(ref, i + 1))
for ng in hypcnts:
max_counts[ng] = max(max_counts.get(ng, 0), refcnts[ng])
clipcnt = dict((ng, min(count, max_counts[ng])) \
for ng, count in hypcnts.items())
clip_count[i] += sum(clipcnt.values())
# accumulate r & c
bestmatch = [1000, 1000]
for ref in refs:
if bestmatch[0] == 0: break
diff = abs(len(ref) - len(hyp))
if diff < bestmatch[0]:
bestmatch[0] = diff
bestmatch[1] = len(ref)
r += bestmatch[1]
c += len(hyp)
# computing bleu score
p0 = 1e-7
bp = 1 if c > r else math.exp(1 - float(r) / float(c))
p_ns = [float(clip_count[i]) / float(count[i] + p0) + p0 \
for i in range(4)]
s = math.fsum(w * math.log(p_n) \
for w, p_n in zip(weights, p_ns) if p_n)
bleu = bp * math.exp(s)
return bleu
def report(func):
@functools.wraps(func)
def wrapper(*args, **kwargs):
res = func(*args, **kwargs)
args[0].metric_dict[func.__name__ + ' '+str(args[2])] = res
return res
return wrapper
class GenericEvaluator:
def __init__(self, result_path):
self.file = open(result_path,'r')
self.meta = []
self.metric_dict = {}
self.entity_dict = {}
filename = result_path.split('/')[-1]
dump_dir = './sheets/' + filename.replace('.csv','.report.txt')
self.dump_file = open(dump_dir,'w')
def _print_dict(self, dic):
for k, v in sorted(dic.items(),key=lambda x:x[0]):
print(k+'\t'+str(v))
@report
def bleu_metric(self,data,type='bleu'):
gen, truth = [],[]
for row in data:
gen.append(row['generated_response'])
truth.append(row['response'])
wrap_generated = [[_] for _ in gen]
wrap_truth = [[_] for _ in truth]
sc = BLEUScorer().score(zip(wrap_generated, wrap_truth))
return sc
def run_metrics(self):
raise ValueError('Please specify the evaluator first, bro')
def read_result_data(self):
while True:
line = self.file.readline()
if 'START_CSV_SECTION' in line:
break
self.meta.append(line)
reader = csv.DictReader(self.file)
data = [_ for _ in reader]
return data
def _extract_constraint(self, z):
z = z.split()
if 'EOS_Z1' not in z:
return set(z).difference(['name', 'address', 'postcode', 'phone', 'area', 'pricerange', 'restaurant',
'restaurants', 'style', 'price', 'food', 'EOS_M'])
else:
idx = z.index('EOS_Z1')
return set(z[:idx]).difference(['name', 'address', 'postcode', 'phone', 'area', 'pricerange', 'restaurant',
'restaurants', 'style', 'price', 'food', 'EOS_M'])
def _extract_request(self, z):
z = z.split()
if 'EOS_Z1' not in z or z[-1] == 'EOS_Z1':
return set()
else:
idx = z.index('EOS_Z1')
return set(z[idx+1:])
def pack_dial(self,data):
dials = {}
for turn in data:
dial_id = int(turn['dial_id'])
if dial_id not in dials:
dials[dial_id] = []
dials[dial_id].append(turn)
return dials
def dump(self):
self.dump_file.writelines(self.meta)
self.dump_file.write('START_REPORT_SECTION\n')
for k,v in self.metric_dict.items():
self.dump_file.write('{}\t{}\n'.format(k,v))
def clean(self,s):
s = s.replace('<go> ', '').replace(' SLOT', '_SLOT')
s = '<GO> ' + s + ' </s>'
for item in self.entity_dict:
# s = s.replace(item, 'VALUE_{}'.format(self.entity_dict[item]))
s = clean_replace(s, item, '{}_SLOT'.format(self.entity_dict[item]))
return s
class CamRestEvaluator(GenericEvaluator):
def __init__(self, result_path):
super().__init__(result_path)
self.entities = []
self.entity_dict = {}
def run_metrics(self):
# # to go over all the OTGY files.
# raw_json = open('./data/CamRest676/CamRest676.json')
# raw_entities = open('./data/CamRest676/CamRestOTGY.json')
# idx = 1
raw_json = open(cfg.data)
raw_entities = open(cfg.entity)
# raw_json = open('../SimDial/1500_data_fixed_'+str(idx)+'/restaurant-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed_'+str(idx)+'/restaurant-MixSpec-1500-OTGY.json')
# # raw_json = open('../SimDial/1500_data_fixed_'+str(idx)+'/movie-MixSpec-1500.json')
# # raw_entities = open('../SimDial/1500_data_fixed_'+str(idx)+'/movie-MixSpec-1500-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed_'+str(idx)+'/rest_pitt-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed_'+str(idx)+'/rest_pitt-MixSpec-1500-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed_'+str(idx)+'/restaurant_style-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed_'+str(idx)+'/restaurant_style-MixSpec-1500-OTGY.json')
raw_data = json.loads(raw_json.read().lower())
raw_entities = json.loads(raw_entities.read().lower())
self.get_entities(raw_entities)
data = self.read_result_data()
for i, row in enumerate(data):
data[i]['response'] = self.clean(data[i]['response'])
data[i]['generated_response'] = self.clean(data[i]['generated_response'])
bleu_score = self.bleu_metric(data,'bleu')
success_f1 = self.success_f1_metric(data, 'success')
match = self.match_metric(data, 'match', raw_data=raw_data)
# #############################
# pdb.set_trace()
# #############################
self._print_dict(self.metric_dict)
return -success_f1[0]
def get_entities(self, entity_data):
for k in entity_data['informable']:
self.entities.extend(entity_data['informable'][k])
for item in entity_data['informable'][k]:
self.entity_dict[item] = k
def run_metrics_maml(self):
# # to go over all the OTGY files.
# # # for the enitites
# raw_entities_path_list = cfg.entity
# raw_entities = []
# for entity_path in raw_entities_path_list:
# raw_entity = open(entity_path)
# raw_entities.append(json.loads(raw_entity.read().lower()))
# # #############################
# # pdb.set_trace()
# # #############################
# self.get_entities_maml(raw_entities)
raw_json = open(cfg.data)
raw_entities = open(cfg.entity)
# raw_json = open('./data/CamRest676/CamRest676.json')
# raw_entities = open('./data/CamRest676/CamRestOTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/r_w_b.json')
# raw_entities = open('../SimDial/1500_data_fixed/r_w_b-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/rest_weat_bus_littlemovie.json')
# raw_entities = open('../SimDial/1500_data_fixed/rest_weat_bus_littlemovie-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/restaurant-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed/restaurant-MixSpec-1500-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/movie-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed/movie-MixSpec-1500-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/movie-MixSpec-150x10.json')
# raw_entities = open('../SimDial/1500_data_fixed/movie-MixSpec-150-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/rest_pitt-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed/rest_pitt-MixSpec-1500-OTGY.json')
# raw_json = open('../SimDial/1500_data_fixed/restaurant_style-MixSpec-1500.json')
# raw_entities = open('../SimDial/1500_data_fixed/restaurant_style-MixSpec-1500-OTGY.json')
raw_data = json.loads(raw_json.read().lower())
raw_entities = json.loads(raw_entities.read().lower())
self.get_entities(raw_entities)
# # # for the data
# raw_json_path_list = cfg.data
# raw_data = []
# # raw_json = open('./data/CamRest676/CamRest676.json')
# for raw_json_path in raw_json_path_list:
# raw_json = open(raw_json_path)
# raw_data += json.loads(raw_json.read().lower())
data = self.read_result_data()
for i, row in enumerate(data):
data[i]['response'] = self.clean(data[i]['response'])
data[i]['generated_response'] = self.clean(data[i]['generated_response'])
bleu_score = self.bleu_metric(data,'bleu')
success_f1 = self.success_f1_metric(data, 'success')
match = self.match_metric(data, 'match', raw_data=raw_data)
# #############################
# pdb.set_trace()
# #############################
self._print_dict(self.metric_dict)
return -success_f1[0]
def get_entities_maml(self, entities_data):
for entity_data in entities_data:
# #############################
# pdb.set_trace()
# #############################
for k in entity_data['informable']:
self.entities.extend(entity_data['informable'][k])
for item in entity_data['informable'][k]:
self.entity_dict[item] = k
def _extract_constraint(self, z):
z = z.split()
if 'EOS_Z1' not in z:
s = set(z)
else:
idx = z.index('EOS_Z1')
s = set(z[:idx])
if 'moderately' in s:
s.discard('moderately')
s.add('moderate')
return s.intersection(self.entities)
###############moderate###############
# if 'EOS_Z1' in z:
# idx = z.index('EOS_Z1')
# z = z[:idx]
# s = set()
# for i in self.entities:
# if i in z:
# s.add(i)
# return s
def _extract_request(self, z):
z = z.split()
return set(z).intersection(['address', 'postcode', 'phone', 'area', 'pricerange','food',
'open', 'price', 'parking',
'duration', 'arrive_in',
'temperature', 'weather_type',
'rating', 'company', 'director'])
@report
def match_metric(self, data, sub='match',raw_data=None):
dials = self.pack_dial(data)
match,total = 0,1e-8
tp, fn, fp = 0.0, 0.0, 0.0
success = 0
count1 = 0
# find out the last placeholder and see whether that is correct
# if no such placeholder, see the final turn, because it can be a yes/no question or scheduling dialogue
for dial_id in dials:
truth_req, gen_req = [], []
dial = dials[dial_id]
gen_bspan, truth_cons, gen_cons = None, None, set()
truth_turn_num = -1
truth_response_req = []
for turn_num,turn in enumerate(dial):
# #############################
# if 'unk' in turn['user']:
# pdb.set_trace()
# #############################
if 'SLOT' in turn['generated_response']:
gen_bspan = turn['generated_bspan']
gen_cons = self._extract_constraint(gen_bspan)
if 'SLOT' in turn['response']:
truth_cons = self._extract_constraint(turn['bspan'])
gen_response_token = turn['generated_response'].split()
response_token = turn['response'].split()
for idx, w in enumerate(gen_response_token):
if w.endswith('SLOT') and w != 'SLOT':
gen_req.append(w.split('_')[0])
if w == 'SLOT' and idx != 0:
gen_req.append(gen_response_token[idx - 1])
for idx, w in enumerate(response_token):
if w.endswith('SLOT') and w != 'SLOT':
truth_response_req.append(w.split('_')[0])
# #############################
# pdb.set_trace()
# #############################
if not gen_cons:
gen_bspan = dial[-1]['generated_bspan']
gen_cons = self._extract_constraint(gen_bspan)
# #############################
# if 'unk' in turn['generated_bspan']:
# pdb.set_trace()
# #############################
if truth_cons:
# #############################
# pdb.set_trace()
# #############################
tp_tmp = len([t for t in gen_cons if t in truth_cons])
fp_tmp = max(0, len(gen_cons) - tp_tmp)
fn_tmp = max(0, len(truth_cons) - tp_tmp)
tp += tp_tmp
fp += fp_tmp
fn += fn_tmp
if gen_cons == truth_cons:
#############################
# if 'unk' in turn['user'] + turn['response'] + turn['generated_response'] + turn['generated_bspan']:
# pdb.set_trace()
#############################
match += 1
# else:
# print(gen_cons, truth_cons)
# #############################
# pdb.set_trace()
# #############################
total += 1
else:
# #############################
# pdb.set_trace()
# #############################
count1 += 1
precision, recall = tp / (tp + fp + 1e-8), tp / (tp + fn + 1e-8)
entity_f1 = 2 * precision * recall / (precision + recall + 1e-8)
print(total, count1)
# print("tp: %f , fn: %f, fp: %f" %(tp,fn,fp))
print("precision: %f; recall: %f" %(precision, recall))
print("entity f1:", entity_f1)
return match / total, entity_f1
@report
def success_f1_metric(self, data, sub='successf1'):
dials = self.pack_dial(data)
tp,fp,fn = 0,0,0
for dial_id in dials:
truth_req, gen_req = set(),set()
dial = dials[dial_id]
for turn_num, turn in enumerate(dial):
gen_response_token = turn['generated_response'].split()
response_token = turn['response'].split()
for idx, w in enumerate(gen_response_token):
if w.endswith('SLOT') and w != 'SLOT':
gen_req.add(w.split('_')[0])
for idx, w in enumerate(response_token):
if w.endswith('SLOT') and w != 'SLOT':
truth_req.add(w.split('_')[0])
gen_req.discard('name')
truth_req.discard('name')
for req in gen_req:
if req in truth_req:
tp += 1
else:
fp += 1
for req in truth_req:
if req not in gen_req:
fn += 1
precision, recall = tp / (tp + fp + 1e-8), tp / (tp + fn + 1e-8)
f1 = 2 * precision * recall / (precision + recall + 1e-8)
return f1, precision, recall
class KvretEvaluator(GenericEvaluator):
def __init__(self, result_path):
super().__init__(result_path)
ent_json = open('./data/kvret/kvret_entities.json')
self.ent_data = json.loads(ent_json.read().lower())
self._get_entity_dict(self.ent_data)
raw_json = open('./data/kvret/kvret_test_public.json')
self.raw_data = json.loads(raw_json.read().lower())
raw_json.close()
def run_metrics(self):
data = self.read_result_data()
for i, row in enumerate(data):
data[i]['response'] = self.clean_by_intent(data[i]['response'],int(data[i]['dial_id']))
data[i]['generated_response'] = self.clean_by_intent(data[i]['generated_response'],int(data[i]['dial_id']))
match_rate = self.match_rate_metric(data, 'match')
bleu_score = self.bleu_metric(data,'bleu')
success_f1 = self.success_f1_metric(data,'success_f1')
self._print_dict(self.metric_dict)
def clean_by_intent(self,s,i):
s = s.replace('<go> ', '').replace(' SLOT', '_SLOT')
s = '<GO> ' + s + ' </s>'
intent = self.raw_data[i]['scenario']['task']['intent']
slot = {
'weather':['weather_attribute','location','weekly_time'],
'navigate':['poi','poi_type','distance','traffic','address'],
'schedule':['event','date','time','party','room','agenda']
}
for item in self.entity_dict:
if self.entity_dict[item] in slot[intent]:
# s = s.replace(item, 'VALUE_{}'.format(self.entity_dict[item]))
s = clean_replace(s, item, '{}_SLOT'.format(self.entity_dict[item]))
return s
def _extract_constraint(self, z):
z = z.split()
if 'EOS_Z1' not in z:
s = set(z)
else:
idx = z.index('EOS_Z1')
s = set(z[:idx])
reqs = ['address', 'traffic', 'poi', 'poi_type', 'distance', 'weather', 'temperature', 'weather_attribute',
'date', 'time', 'location', 'event', 'agenda', 'party', 'room', 'weekly_time', 'forecast']
informable = {
'weather': ['date','location','weather_attribute'],
'navigate': ['poi_type','distance'],
'schedule': ['event', 'date', 'time', 'agenda', 'party', 'room']
}
infs = []
for v in informable.values():
infs.extend(v)
junk = ['good','great','quickest','shortest','route','week','fastest','nearest','next','closest','way','mile',
'activity','restaurant','appointment' ]
s = s.difference(junk).difference(en_sws).difference(reqs)
res = set()
for item in s:
if item in junk:
continue
flg = False
for canon_ent in sorted(list(self.entity_dict.keys())):
if self.entity_dict[canon_ent] in infs:
if similar(item, canon_ent):
flg = True
junk.extend(canon_ent.split())
res.add(canon_ent)
if flg:
break
return res
def constraint_same(self, truth_cons, gen_cons):
if not truth_cons and not gen_cons:
return True
if not truth_cons or not gen_cons:
return False
return setsim(gen_cons, truth_cons)
def _get_entity_dict(self, entity_data):
entity_dict = {}
for k in entity_data:
if type(entity_data[k][0]) is str:
for entity in entity_data[k]:
entity = self._lemmatize(self._tokenize(entity))
entity_dict[entity] = k
if k in ['event','poi_type']:
entity_dict[entity.split()[0]] = k
elif type(entity_data[k][0]) is dict:
for entity_entry in entity_data[k]:
for entity_type, entity in entity_entry.items():
entity_type = 'poi_type' if entity_type == 'type' else entity_type
entity = self._lemmatize(self._tokenize(entity))
entity_dict[entity] = entity_type
if entity_type in ['event', 'poi_type']:
entity_dict[entity.split()[0]] = entity_type
self.entity_dict = entity_dict
@report
def match_rate_metric(self, data, sub='match',bspans='./data/kvret/test.bspan.pkl'):
dials = self.pack_dial(data)
match,total = 0,1e-8
#bspan_data = pickle.load(open(bspans,'rb'))
# find out the last placeholder and see whether that is correct
# if no such placeholder, see the final turn, because it can be a yes/no question or scheduling conversation
for dial_id in dials:
dial = dials[dial_id]
gen_bspan, truth_cons, gen_cons = None, None, set()
truth_turn_num = -1
for turn_num,turn in enumerate(dial):
if 'SLOT' in turn['generated_response']:
gen_bspan = turn['generated_bspan']
gen_cons = self._extract_constraint(gen_bspan)
if 'SLOT' in turn['response']:
truth_cons = self._extract_constraint(turn['bspan'])
# KVRET dataset includes "scheduling" (so often no SLOT decoded in ground truth)
if not truth_cons:
truth_bspan = dial[-1]['bspan']
truth_cons = self._extract_constraint(truth_bspan)
if not gen_cons:
gen_bspan = dial[-1]['generated_bspan']
gen_cons = self._extract_constraint(gen_bspan)
if truth_cons:
if self.constraint_same(gen_cons, truth_cons):
match += 1
#print(gen_cons, truth_cons, '+')
else:
print(gen_cons, truth_cons, '-')
total += 1
return match / total
def _tokenize(self, sent):
return ' '.join(word_tokenize(sent))
def _lemmatize(self, sent):
words = [wn.lemmatize(_) for _ in sent.split()]
#for idx,w in enumerate(words):
# if w !=
return ' '.join(words)
@report
def success_f1_metric(self, data, sub='successf1'):
dials = self.pack_dial(data)
tp,fp,fn = 0,0,0
for dial_id in dials:
truth_req, gen_req = set(),set()
dial = dials[dial_id]
for turn_num, turn in enumerate(dial):
gen_response_token = turn['generated_response'].split()
response_token = turn['response'].split()
for idx, w in enumerate(gen_response_token):
if w.endswith('SLOT') and w != 'SLOT':
gen_req.add(w.split('_')[0])
for idx, w in enumerate(response_token):
if w.endswith('SLOT') and w != 'SLOT':
truth_req.add(w.split('_')[0])
gen_req.discard('name')
truth_req.discard('name')
for req in gen_req:
if req in truth_req:
tp += 1
else:
fp += 1
for req in truth_req:
if req not in gen_req:
fn += 1
precision, recall = tp / (tp + fp + 1e-8), tp / (tp + fn + 1e-8)
f1 = 2 * precision * recall / (precision + recall + 1e-8)
return f1
def metric_handler():
parser = argparse.ArgumentParser()
parser.add_argument('-file')
parser.add_argument('-type')
args = parser.parse_args()
ev_class = None
if args.type == 'camrest':
ev_class = CamRestEvaluator
elif args.type == 'kvret':
ev_class = KvretEvaluator
ev = ev_class(args.file)
ev.run_metrics()
ev.dump()
if __name__ == '__main__':
metric_handler()