google-research

1
# coding=utf-8
2
# Copyright 2024 The Google Research Authors.
3
#
4
# Licensed under the Apache License, Version 2.0 (the "License");
5
# you may not use this file except in compliance with the License.
6
# You may obtain a copy of the License at
7
#
8
#     http://www.apache.org/licenses/LICENSE-2.0
9
#
10
# Unless required by applicable law or agreed to in writing, software
11
# distributed under the License is distributed on an "AS IS" BASIS,
12
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
# See the License for the specific language governing permissions and
14
# limitations under the License.
15

16
"""Tests for schema_guided_dst.baseline.data_utils."""
17

18
from __future__ import absolute_import
19
from __future__ import division
20
from __future__ import print_function
21

22
import os
23

24
from absl.testing import absltest
25

26
from schema_guided_dst.baseline import config
27
from schema_guided_dst.baseline import data_utils
28

29
_VOCAB_FILE = os.path.join(os.path.dirname(os.path.abspath(__file__)),
30
                           'test_data/bert_vocab.txt')
31
_TEST_DATA_DIR = os.path.dirname(_VOCAB_FILE)
32
_DO_LOWER_CASE = True
33
_DATASET = 'train'
34

35

36
class Dstc8DataProcessorTest(absltest.TestCase):
37
  """Tests for Dstc8DataProcessor."""
38

39
  def setUp(self):
40
    self._processor = data_utils.Dstc8DataProcessor(
41
        dstc8_data_dir=_TEST_DATA_DIR,
42
        dataset_config=config.DatasetConfig(
43
            file_ranges={
44
                'train': range(1),
45
                'dev': None,
46
                'test': None
47
            },
48
            max_num_cat_slot=6,
49
            max_num_noncat_slot=6,
50
            max_num_value_per_cat_slot=4,
51
            max_num_intent=2),
52
        vocab_file=_VOCAB_FILE,
53
        do_lower_case=_DO_LOWER_CASE)
54
    super(Dstc8DataProcessorTest, self).setUp()
55

56
  def test_tokenizer(self):
57
    # Test normal sentence.
58
    test_utt_1 = 'Watch, Hellboy?'
59
    utt_1_tokens, utt_1_aligns, utt_1_inv_alignments = (
60
        self._processor._tokenize(test_utt_1))
61
    expected_utt_1_tokens = ['watch', ',', 'hell', '##boy', '?']
62
    expected_utt_1_aligns = {0: 0, 4: 0, 5: 1, 7: 2, 13: 3, 14: 4}
63
    expected_utt_1_inv_alignments = [(0, 4), (5, 5), (7, 13), (7, 13), (14, 14)]
64
    self.assertEqual(utt_1_tokens, expected_utt_1_tokens)
65
    self.assertEqual(utt_1_aligns, expected_utt_1_aligns)
66
    self.assertEqual(utt_1_inv_alignments, expected_utt_1_inv_alignments)
67

68
    # Test extra spaces in the utterance.
69
    test_utt_2 = 'Extra  ,  spaces'
70
    utt_2_tokens, utt_2_aligns, utt_2_inv_alignments = (
71
        self._processor._tokenize(test_utt_2))
72
    expected_utt_1_inv_alignments = [(0, 4), (5, 5), (7, 13), (7, 13), (14, 14)]
73
    self.assertEqual(utt_2_tokens, ['extra', ',', 'spaces'])
74
    self.assertEqual(utt_2_aligns, {0: 0, 4: 0, 7: 1, 10: 2, 15: 2})
75
    self.assertEqual(utt_2_inv_alignments, [(0, 4), (7, 7), (10, 15)])
76

77
    # Test # appearing in the string.
78
    test_utt_3 = 'Extra## ##abc'
79
    utt_3_tokens, utt_3_aligns, utt_3_inv_alignments = (
80
        self._processor._tokenize(test_utt_3))
81
    self.assertEqual(utt_3_tokens,
82
                     ['extra', '#', '#', '#', '#', 'a', '##b', '##c'])
83
    self.assertEqual(utt_3_aligns, {
84
        0: 0,
85
        4: 0,
86
        5: 1,
87
        6: 2,
88
        8: 3,
89
        9: 4,
90
        10: 5,
91
        12: 7
92
    })
93
    self.assertEqual(utt_3_inv_alignments, [(0, 4), (5, 5), (6, 6), (8, 8),
94
                                            (9, 9), (10, 12), (10, 12),
95
                                            (10, 12)])
96

97
  def test_get_dialog_examples(self):
98
    examples = self._processor.get_dialog_examples(_DATASET)
99
    # Check that the summary of all the turns are correct.
100
    expected_summaries = [
101
        {
102
            'utt_tok_mask_pairs': [('[CLS]', 0), ('[SEP]', 0),
103
                                   ('i', 1), ("'", 1), ('m', 1), ('looking', 1),
104
                                   ('for', 1), ('apartments', 1), ('.', 1),
105
                                   ('[SEP]', 1)],
106
            'utt_len': 10,
107
            'num_categorical_slots': 4,
108
            'num_categorical_slot_values': [2, 4, 4, 2, 0, 0],
109
            'num_noncategorical_slots': 3,
110
            'service_name': 'Homes_1',
111
            'active_intent': 'FindApartment',
112
            'slot_values_in_state': {}
113
        },
114
        {
115
            'utt_tok_mask_pairs': [('[CLS]', 0), ('which', 0), ('area', 0),
116
                                   ('are', 0), ('you', 0), ('looking', 0),
117
                                   ('in', 0), ('?', 0), ('[SEP]', 0), ('i', 1),
118
                                   ('want', 1), ('an', 1), ('apartment', 1),
119
                                   ('in', 1), ('sa', 1), ('##n', 1), ('j', 1),
120
                                   ('##ose', 1), ('.', 1), ('[SEP]', 1)],
121
            'utt_len': 20,
122
            'num_categorical_slots': 4,
123
            'num_categorical_slot_values': [2, 4, 4, 2, 0, 0],
124
            'num_noncategorical_slots': 3,
125
            'service_name': 'Homes_1',
126
            'active_intent': 'FindApartment',
127
            'slot_values_in_state': {
128
                'area': 'san jose'
129
            }
130
        },
131
        {
132
            'utt_tok_mask_pairs': [('[CLS]', 0), ('how', 0), ('many', 0),
133
                                   ('bedrooms', 0), ('do', 0), ('you', 0),
134
                                   ('want', 0), ('?', 0), ('[SEP]', 0),
135
                                   ('2', 1), ('bedrooms', 1), (',', 1),
136
                                   ('please', 1), ('.', 1), ('[SEP]', 1)],
137
            'utt_len': 15,
138
            'num_categorical_slots': 4,
139
            'num_categorical_slot_values': [2, 4, 4, 2, 0, 0],
140
            'num_noncategorical_slots': 3,
141
            'service_name': 'Homes_1',
142
            'active_intent': 'FindApartment',
143
            'slot_values_in_state': {
144
                'number_of_beds': '2'
145
            }
146
        },
147
        {
148
            'utt_tok_mask_pairs': [
149
                ('[CLS]', 0), ('there', 0), ("'", 0), ('s', 0), ('a', 0),
150
                ('nice', 0), ('property', 0), ('called', 0), ('a', 0),
151
                ('##ege', 0), ('##na', 0), ('at', 0), ('129', 0), ('##0', 0),
152
                ('sa', 0), ('##n', 0), ('to', 0), ('##mas', 0), ('a', 0),
153
                ('##quin', 0), ('##o', 0), ('road', 0), ('.', 0), ('it', 0),
154
                ('has', 0), ('2', 0), ('bedrooms', 0), (',', 0), ('1', 0),
155
                ('bath', 0), (',', 0), ('and', 0), ('rent', 0), ('##s', 0),
156
                ('for', 0), ('$', 0), ('2', 0), (',', 0), ('650', 0), ('a', 0),
157
                ('month', 0), ('.', 0), ('[SEP]', 0), ('can', 1), ('you', 1),
158
                ('find', 1), ('me', 1), ('a', 1), ('three', 1), ('bedroom', 1),
159
                ('apartment', 1), ('in', 1), ('liver', 1), ('##more', 1),
160
                ('?', 1), ('[SEP]', 1)
161
            ],
162
            'utt_len': 56,
163
            'num_categorical_slots': 4,
164
            'num_categorical_slot_values': [2, 4, 4, 2, 0, 0],
165
            'num_noncategorical_slots': 3,
166
            'service_name': 'Homes_1',
167
            'active_intent': 'FindApartment',
168
            'slot_values_in_state': {
169
                'number_of_beds': '3',
170
                'area': 'livermore'
171
            }
172
        },
173
        {
174
            'utt_tok_mask_pairs': [
175
                ('[CLS]', 0), ('there', 0), ("'", 0), ('s', 0), ('a', 0),
176
                ('##cacia', 0), ('capital', 0), ('co', 0), ('##r', 0), ('-', 0),
177
                ('iron', 0), ('##wood', 0), ('a', 0), ('##p', 0), ('at', 0),
178
                ('56', 0), ('##43', 0), ('ch', 0), ('##ar', 0), ('##lot', 0),
179
                ('##te', 0), ('way', 0), ('.', 0), ('it', 0), ('has', 0),
180
                ('3', 0), ('bedrooms', 0), (',', 0), ('3', 0), ('baths', 0),
181
                (',', 0), ('and', 0), ('rent', 0), ('##s', 0), ('for', 0),
182
                ('$', 0), ('4', 0), (',', 0), ('05', 0), ('##0', 0), ('a', 0),
183
                ('month', 0), ('.', 0), ('[SEP]', 0), ('that', 1), ('one', 1),
184
                ('sounds', 1), ('good', 1), ('.', 1), ('thanks', 1), (',', 1),
185
                ('that', 1), ("'", 1), ('s', 1), ('all', 1), ('i', 1),
186
                ('need', 1), ('.', 1), ('[SEP]', 1)
187
            ],
188
            'utt_len': 59,
189
            'num_categorical_slots': 4,
190
            'num_categorical_slot_values': [2, 4, 4, 2, 0, 0],
191
            'num_noncategorical_slots': 3,
192
            'service_name': 'Homes_1',
193
            'active_intent': 'FindApartment',
194
            'slot_values_in_state': {
195
                'property_name': 'acacia capital cor - ironwood ap'
196
            }
197
        },
198
    ]
199
    for example, gold in zip(examples, expected_summaries):
200
      self.assertEqual(example.readable_summary, gold)
201

202

203
if __name__ == '__main__':
204
  absltest.main()
205