/
test_evaluation.py
765 lines (655 loc) · 23 KB
/
test_evaluation.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
import asyncio
import logging
import pytest
import rasa.utils.io
from rasa.test import compare_nlu_models
from rasa.nlu.extractors import EntityExtractor
from rasa.nlu.extractors.mitie_entity_extractor import MitieEntityExtractor
from rasa.nlu.extractors.spacy_entity_extractor import SpacyEntityExtractor
from rasa.nlu.model import Interpreter
from rasa.nlu.test import (
is_token_within_entity,
do_entities_overlap,
merge_labels,
remove_empty_intent_examples,
remove_empty_response_examples,
get_entity_extractors,
remove_pretrained_extractors,
drop_intents_below_freq,
cross_validate,
run_evaluation,
substitute_labels,
IntentEvaluationResult,
EntityEvaluationResult,
ResponseSelectionEvaluationResult,
evaluate_intents,
evaluate_entities,
evaluate_response_selections,
get_unique_labels,
get_evaluation_metrics,
NO_ENTITY,
collect_successful_entity_predictions,
collect_incorrect_entity_predictions,
)
from rasa.nlu.test import does_token_cross_borders
from rasa.nlu.test import align_entity_predictions
from rasa.nlu.test import determine_intersection
from rasa.nlu.test import determine_token_labels
from rasa.nlu.config import RasaNLUModelConfig
from rasa.nlu.tokenizers import Token
from rasa.nlu import utils
import json
import os
from rasa.nlu import training_data, config
from tests.nlu import utilities
from tests.nlu.conftest import DEFAULT_DATA_PATH, NLU_DEFAULT_CONFIG_PATH
@pytest.fixture(scope="session")
def loop():
loop = asyncio.new_event_loop()
asyncio.set_event_loop(loop)
loop = rasa.utils.io.enable_async_loop_debugging(loop)
yield loop
loop.close()
@pytest.fixture(scope="session")
async def pretrained_interpreter(component_builder, tmpdir_factory):
conf = RasaNLUModelConfig(
{
"pipeline": [
{"name": "SpacyNLP"},
{"name": "SpacyEntityExtractor"},
{"name": "DucklingHTTPExtractor"},
]
}
)
return await utilities.interpreter_for(
component_builder,
data="./data/examples/rasa/demo-rasa.json",
path=tmpdir_factory.mktemp("projects").strpath,
config=conf,
)
# Chinese Example
# "对面食过敏" -> To be allergic to wheat-based food
CH_wrong_segmentation = [
Token("对面", 0),
Token("食", 2),
Token("过敏", 3), # opposite, food, allergy
]
CH_correct_segmentation = [
Token("对", 0),
Token("面食", 1),
Token("过敏", 3), # towards, wheat-based food, allergy
]
CH_wrong_entity = {"start": 0, "end": 2, "value": "对面", "entity": "direction"}
CH_correct_entity = {"start": 1, "end": 3, "value": "面食", "entity": "food_type"}
# EN example
# "Hey Robot, I would like to eat pizza near Alexanderplatz tonight"
EN_indices = [0, 4, 9, 11, 13, 19, 24, 27, 31, 37, 42, 57]
EN_tokens = [
"Hey",
"Robot",
",",
"I",
"would",
"like",
"to",
"eat",
"pizza",
"near",
"Alexanderplatz",
"tonight",
]
EN_tokens = [Token(t, i) for t, i in zip(EN_tokens, EN_indices)]
EN_targets = [
{"start": 31, "end": 36, "value": "pizza", "entity": "food"},
{"start": 37, "end": 56, "value": "near Alexanderplatz", "entity": "location"},
{"start": 57, "end": 64, "value": "tonight", "entity": "datetime"},
]
EN_predicted = [
{
"start": 4,
"end": 9,
"value": "Robot",
"entity": "person",
"extractor": "EntityExtractorA",
},
{
"start": 31,
"end": 36,
"value": "pizza",
"entity": "food",
"extractor": "EntityExtractorA",
},
{
"start": 42,
"end": 56,
"value": "Alexanderplatz",
"entity": "location",
"extractor": "EntityExtractorA",
},
{
"start": 42,
"end": 64,
"value": "Alexanderplatz tonight",
"entity": "movie",
"extractor": "EntityExtractorB",
},
]
EN_entity_result = EntityEvaluationResult(
EN_targets, EN_predicted, EN_tokens, " ".join([t.text for t in EN_tokens])
)
EN_entity_result_no_tokens = EntityEvaluationResult(EN_targets, EN_predicted, [], "")
def test_token_entity_intersection():
# included
intsec = determine_intersection(CH_correct_segmentation[1], CH_correct_entity)
assert intsec == len(CH_correct_segmentation[1].text)
# completely outside
intsec = determine_intersection(CH_correct_segmentation[2], CH_correct_entity)
assert intsec == 0
# border crossing
intsec = determine_intersection(CH_correct_segmentation[1], CH_wrong_entity)
assert intsec == 1
def test_token_entity_boundaries():
# smaller and included
assert is_token_within_entity(CH_wrong_segmentation[1], CH_correct_entity)
assert not does_token_cross_borders(CH_wrong_segmentation[1], CH_correct_entity)
# exact match
assert is_token_within_entity(CH_correct_segmentation[1], CH_correct_entity)
assert not does_token_cross_borders(CH_correct_segmentation[1], CH_correct_entity)
# completely outside
assert not is_token_within_entity(CH_correct_segmentation[0], CH_correct_entity)
assert not does_token_cross_borders(CH_correct_segmentation[0], CH_correct_entity)
# border crossing
assert not is_token_within_entity(CH_wrong_segmentation[0], CH_correct_entity)
assert does_token_cross_borders(CH_wrong_segmentation[0], CH_correct_entity)
def test_entity_overlap():
assert do_entities_overlap([CH_correct_entity, CH_wrong_entity])
assert not do_entities_overlap(EN_targets)
def test_determine_token_labels_throws_error():
with pytest.raises(ValueError):
determine_token_labels(
CH_correct_segmentation[0],
[CH_correct_entity, CH_wrong_entity],
["CRFEntityExtractor"],
)
def test_determine_token_labels_no_extractors():
with pytest.raises(ValueError):
determine_token_labels(
CH_correct_segmentation[0], [CH_correct_entity, CH_wrong_entity], None
)
def test_determine_token_labels_no_extractors_no_overlap():
determine_token_labels(CH_correct_segmentation[0], EN_targets, None)
def test_determine_token_labels_with_extractors():
determine_token_labels(
CH_correct_segmentation[0],
[CH_correct_entity, CH_wrong_entity],
[SpacyEntityExtractor.name, MitieEntityExtractor.name],
)
def test_label_merging():
aligned_predictions = [
{
"target_labels": ["O", "O"],
"extractor_labels": {"EntityExtractorA": ["O", "O"]},
},
{
"target_labels": ["LOC", "O", "O"],
"extractor_labels": {"EntityExtractorA": ["O", "O", "O"]},
},
]
assert all(merge_labels(aligned_predictions) == ["O", "O", "LOC", "O", "O"])
assert all(
merge_labels(aligned_predictions, "EntityExtractorA")
== ["O", "O", "O", "O", "O"]
)
def test_drop_intents_below_freq():
td = training_data.load_data("data/examples/rasa/demo-rasa.json")
clean_td = drop_intents_below_freq(td, 0)
assert clean_td.intents == {
"affirm",
"goodbye",
"greet",
"restaurant_search",
"chitchat",
}
clean_td = drop_intents_below_freq(td, 10)
assert clean_td.intents == {"affirm", "restaurant_search"}
def test_run_evaluation(unpacked_trained_moodbot_path):
data = DEFAULT_DATA_PATH
result = run_evaluation(
data, os.path.join(unpacked_trained_moodbot_path, "nlu"), errors=None
)
assert result.get("intent_evaluation")
assert result.get("entity_evaluation").get("CRFEntityExtractor")
def test_run_cv_evaluation():
td = training_data.load_data("data/examples/rasa/demo-rasa.json")
nlu_config = config.load("sample_configs/config_pretrained_embeddings_spacy.yml")
n_folds = 2
intent_results, entity_results = cross_validate(td, n_folds, nlu_config)
assert len(intent_results.train["Accuracy"]) == n_folds
assert len(intent_results.train["Precision"]) == n_folds
assert len(intent_results.train["F1-score"]) == n_folds
assert len(intent_results.test["Accuracy"]) == n_folds
assert len(intent_results.test["Precision"]) == n_folds
assert len(intent_results.test["F1-score"]) == n_folds
assert len(entity_results.train["CRFEntityExtractor"]["Accuracy"]) == n_folds
assert len(entity_results.train["CRFEntityExtractor"]["Precision"]) == n_folds
assert len(entity_results.train["CRFEntityExtractor"]["F1-score"]) == n_folds
assert len(entity_results.test["CRFEntityExtractor"]["Accuracy"]) == n_folds
assert len(entity_results.test["CRFEntityExtractor"]["Precision"]) == n_folds
assert len(entity_results.test["CRFEntityExtractor"]["F1-score"]) == n_folds
def test_intent_evaluation_report(tmpdir_factory):
path = tmpdir_factory.mktemp("evaluation").strpath
report_folder = os.path.join(path, "reports")
report_filename = os.path.join(report_folder, "intent_report.json")
rasa.utils.io.create_directory(report_folder)
intent_results = [
IntentEvaluationResult("", "restaurant_search", "I am hungry", 0.12345),
IntentEvaluationResult("greet", "greet", "hello", 0.98765),
]
result = evaluate_intents(
intent_results,
report_folder,
successes=False,
errors=False,
confmat_filename=None,
intent_hist_filename=None,
)
report = json.loads(rasa.utils.io.read_file(report_filename))
greet_results = {"precision": 1.0, "recall": 1.0, "f1-score": 1.0, "support": 1}
prediction = {
"text": "hello",
"intent": "greet",
"predicted": "greet",
"confidence": 0.98765,
}
assert len(report.keys()) == 4
assert report["greet"] == greet_results
assert result["predictions"][0] == prediction
def test_response_evaluation_report(tmpdir_factory):
path = tmpdir_factory.mktemp("evaluation").strpath
report_folder = os.path.join(path, "reports")
report_filename = os.path.join(report_folder, "response_selection_report.json")
rasa.utils.io.create_directory(report_folder)
response_results = [
ResponseSelectionEvaluationResult(
"chitchat",
"It's sunny in Berlin",
"It's sunny in Berlin",
"What's the weather",
0.65432,
),
ResponseSelectionEvaluationResult(
"chitchat",
"My name is Mr.bot",
"My name is Mr.bot",
"What's your name?",
0.98765,
),
]
result = evaluate_response_selections(response_results, report_folder)
report = json.loads(rasa.utils.io.read_file(report_filename))
name_query_results = {
"precision": 1.0,
"recall": 1.0,
"f1-score": 1.0,
"support": 1,
}
prediction = {
"text": "What's your name?",
"intent_target": "chitchat",
"response_target": "My name is Mr.bot",
"response_predicted": "My name is Mr.bot",
"confidence": 0.98765,
}
assert len(report.keys()) == 5
assert report["My name is Mr.bot"] == name_query_results
assert result["predictions"][1] == prediction
def test_entity_evaluation_report(tmpdir_factory):
class EntityExtractorA(EntityExtractor):
provides = ["entities"]
def __init__(self, component_config=None) -> None:
super().__init__(component_config)
class EntityExtractorB(EntityExtractor):
provides = ["entities"]
def __init__(self, component_config=None) -> None:
super().__init__(component_config)
path = tmpdir_factory.mktemp("evaluation").strpath
report_folder = os.path.join(path, "reports")
report_filename_a = os.path.join(report_folder, "EntityExtractorA_report.json")
report_filename_b = os.path.join(report_folder, "EntityExtractorB_report.json")
rasa.utils.io.create_directory(report_folder)
mock_interpreter = Interpreter(
[
EntityExtractorA({"provides": ["entities"]}),
EntityExtractorB({"provides": ["entities"]}),
],
None,
)
extractors = get_entity_extractors(mock_interpreter)
result = evaluate_entities([EN_entity_result], extractors, report_folder)
report_a = json.loads(rasa.utils.io.read_file(report_filename_a))
report_b = json.loads(rasa.utils.io.read_file(report_filename_b))
assert len(report_a) == 6
assert report_a["datetime"]["support"] == 1.0
assert report_b["macro avg"]["recall"] == 0.0
assert report_a["macro avg"]["recall"] == 0.5
assert result["EntityExtractorA"]["accuracy"] == 0.75
def test_empty_intent_removal():
intent_results = [
IntentEvaluationResult("", "restaurant_search", "I am hungry", 0.12345),
IntentEvaluationResult("greet", "greet", "hello", 0.98765),
]
intent_results = remove_empty_intent_examples(intent_results)
assert len(intent_results) == 1
assert intent_results[0].intent_target == "greet"
assert intent_results[0].intent_prediction == "greet"
assert intent_results[0].confidence == 0.98765
assert intent_results[0].message == "hello"
def test_empty_response_removal():
response_results = [
ResponseSelectionEvaluationResult(
"chitchat", None, "It's sunny in Berlin", "What's the weather", 0.65432
),
ResponseSelectionEvaluationResult(
"chitchat",
"My name is Mr.bot",
"My name is Mr.bot",
"What's your name?",
0.98765,
),
]
response_results = remove_empty_response_examples(response_results)
assert len(response_results) == 1
assert response_results[0].intent_target == "chitchat"
assert response_results[0].response_target == "My name is Mr.bot"
assert response_results[0].response_prediction == "My name is Mr.bot"
assert response_results[0].confidence == 0.98765
assert response_results[0].message == "What's your name?"
def test_evaluate_entities_cv_empty_tokens():
mock_extractors = ["EntityExtractorA", "EntityExtractorB"]
result = align_entity_predictions(EN_entity_result_no_tokens, mock_extractors)
assert result == {
"target_labels": [],
"extractor_labels": {"EntityExtractorA": [], "EntityExtractorB": []},
}, "Wrong entity prediction alignment"
def test_evaluate_entities_cv():
mock_extractors = ["EntityExtractorA", "EntityExtractorB"]
result = align_entity_predictions(EN_entity_result, mock_extractors)
assert result == {
"target_labels": [
"O",
"O",
"O",
"O",
"O",
"O",
"O",
"O",
"food",
"location",
"location",
"datetime",
],
"extractor_labels": {
"EntityExtractorA": [
"O",
"person",
"O",
"O",
"O",
"O",
"O",
"O",
"food",
"O",
"location",
"O",
],
"EntityExtractorB": [
"O",
"O",
"O",
"O",
"O",
"O",
"O",
"O",
"O",
"O",
"movie",
"movie",
],
},
}, "Wrong entity prediction alignment"
def test_get_entity_extractors(pretrained_interpreter):
assert get_entity_extractors(pretrained_interpreter) == {
"SpacyEntityExtractor",
"DucklingHTTPExtractor",
}
def test_remove_pretrained_extractors(pretrained_interpreter):
target_components_names = ["SpacyNLP"]
filtered_pipeline = remove_pretrained_extractors(pretrained_interpreter.pipeline)
filtered_components_names = [c.name for c in filtered_pipeline]
assert filtered_components_names == target_components_names
def test_label_replacement():
original_labels = ["O", "location"]
target_labels = ["no_entity", "location"]
assert substitute_labels(original_labels, "O", "no_entity") == target_labels
@pytest.mark.parametrize(
"targets,exclude_label,expected",
[
(
["no_entity", "location", "location", "location", "person"],
NO_ENTITY,
["location", "person"],
),
(
["no_entity", "location", "location", "location", "person"],
None,
["no_entity", "location", "person"],
),
(["no_entity"], NO_ENTITY, []),
(["location", "location", "location"], NO_ENTITY, ["location"]),
([], None, []),
],
)
def test_get_label_set(targets, exclude_label, expected):
actual = get_unique_labels(targets, exclude_label)
assert set(expected) == set(actual)
@pytest.mark.parametrize(
"targets,predictions,expected_precision,expected_fscore,expected_accuracy",
[
(
["no_entity", "location", "no_entity", "location", "no_entity"],
["no_entity", "location", "no_entity", "no_entity", "person"],
1.0,
0.6666666666666666,
3 / 5,
),
(
["no_entity", "no_entity", "no_entity", "no_entity", "person"],
["no_entity", "no_entity", "no_entity", "no_entity", "no_entity"],
0.0,
0.0,
4 / 5,
),
],
)
def test_get_evaluation_metrics(
targets, predictions, expected_precision, expected_fscore, expected_accuracy
):
report, precision, f1, accuracy = get_evaluation_metrics(
targets, predictions, True, exclude_label=NO_ENTITY
)
assert f1 == expected_fscore
assert precision == expected_precision
assert accuracy == expected_accuracy
assert NO_ENTITY not in report
def test_nlu_comparison(tmpdir):
configs = [
NLU_DEFAULT_CONFIG_PATH,
"sample_configs/config_supervised_embeddings.yml",
]
output = tmpdir.strpath
compare_nlu_models(
configs, DEFAULT_DATA_PATH, output, runs=2, exclusion_percentages=[50, 80]
)
assert set(os.listdir(output)) == {
"run_1",
"run_2",
"results.json",
"nlu_model_comparison_graph.pdf",
}
run_1_path = os.path.join(output, "run_1")
assert set(os.listdir(run_1_path)) == {"50%_exclusion", "80%_exclusion", "test.md"}
@pytest.mark.parametrize(
"entity_results,targets,predictions,successes,errors",
[
(
[
EntityEvaluationResult(
entity_targets=[
{
"start": 17,
"end": 24,
"value": "Italian",
"entity": "cuisine",
}
],
entity_predictions=[
{
"start": 17,
"end": 24,
"value": "Italian",
"entity": "cuisine",
}
],
tokens=[
"I",
"want",
"to",
"book",
"an",
"Italian",
"restaurant",
".",
],
message="I want to book an Italian restaurant.",
),
EntityEvaluationResult(
entity_targets=[
{
"start": 8,
"end": 15,
"value": "Mexican",
"entity": "cuisine",
},
{
"start": 31,
"end": 32,
"value": "4",
"entity": "number_people",
},
],
entity_predictions=[],
tokens=[
"Book",
"an",
"Mexican",
"restaurant",
"for",
"4",
"people",
".",
],
message="Book an Mexican restaurant for 4 people.",
),
],
[
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
"cuisine",
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
"cuisine",
NO_ENTITY,
NO_ENTITY,
"number_people",
NO_ENTITY,
NO_ENTITY,
],
[
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
"cuisine",
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
NO_ENTITY,
],
[
{
"text": "I want to book an Italian restaurant.",
"entities": [
{
"start": 17,
"end": 24,
"value": "Italian",
"entity": "cuisine",
}
],
"predicted_entities": [
{
"start": 17,
"end": 24,
"value": "Italian",
"entity": "cuisine",
}
],
}
],
[
{
"text": "Book an Mexican restaurant for 4 people.",
"entities": [
{
"start": 8,
"end": 15,
"value": "Mexican",
"entity": "cuisine",
},
{
"start": 31,
"end": 32,
"value": "4",
"entity": "number_people",
},
],
"predicted_entities": [],
}
],
)
],
)
def test_collect_entity_predictions(
entity_results, targets, predictions, successes, errors
):
actual = collect_successful_entity_predictions(entity_results, targets, predictions)
assert len(successes) == len(actual)
assert successes == actual
actual = collect_incorrect_entity_predictions(entity_results, targets, predictions)
assert len(errors) == len(actual)
assert errors == actual