File size: 6,761 Bytes
d916065
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""

Tests for IBM Model 5 training methods

"""

import unittest
from collections import defaultdict

from nltk.translate import AlignedSent, IBMModel, IBMModel4, IBMModel5
from nltk.translate.ibm_model import AlignmentInfo


class TestIBMModel5(unittest.TestCase):
    def test_set_uniform_vacancy_probabilities_of_max_displacements(self):
        # arrange
        src_classes = {"schinken": 0, "eier": 0, "spam": 1}
        trg_classes = {"ham": 0, "eggs": 1, "spam": 2}
        corpus = [
            AlignedSent(["ham", "eggs"], ["schinken", "schinken", "eier"]),
            AlignedSent(["spam", "spam", "spam", "spam"], ["spam", "spam"]),
        ]
        model5 = IBMModel5(corpus, 0, src_classes, trg_classes)

        # act
        model5.set_uniform_probabilities(corpus)

        # assert
        # number of vacancy difference values =
        #     2 * number of words in longest target sentence
        expected_prob = 1.0 / (2 * 4)

        # examine the boundary values for (dv, max_v, trg_class)
        self.assertEqual(model5.head_vacancy_table[4][4][0], expected_prob)
        self.assertEqual(model5.head_vacancy_table[-3][1][2], expected_prob)
        self.assertEqual(model5.non_head_vacancy_table[4][4][0], expected_prob)
        self.assertEqual(model5.non_head_vacancy_table[-3][1][2], expected_prob)

    def test_set_uniform_vacancy_probabilities_of_non_domain_values(self):
        # arrange
        src_classes = {"schinken": 0, "eier": 0, "spam": 1}
        trg_classes = {"ham": 0, "eggs": 1, "spam": 2}
        corpus = [
            AlignedSent(["ham", "eggs"], ["schinken", "schinken", "eier"]),
            AlignedSent(["spam", "spam", "spam", "spam"], ["spam", "spam"]),
        ]
        model5 = IBMModel5(corpus, 0, src_classes, trg_classes)

        # act
        model5.set_uniform_probabilities(corpus)

        # assert
        # examine dv and max_v values that are not in the training data domain
        self.assertEqual(model5.head_vacancy_table[5][4][0], IBMModel.MIN_PROB)
        self.assertEqual(model5.head_vacancy_table[-4][1][2], IBMModel.MIN_PROB)
        self.assertEqual(model5.head_vacancy_table[4][0][0], IBMModel.MIN_PROB)
        self.assertEqual(model5.non_head_vacancy_table[5][4][0], IBMModel.MIN_PROB)
        self.assertEqual(model5.non_head_vacancy_table[-4][1][2], IBMModel.MIN_PROB)

    def test_prob_t_a_given_s(self):
        # arrange
        src_sentence = ["ich", "esse", "ja", "gern", "räucherschinken"]
        trg_sentence = ["i", "love", "to", "eat", "smoked", "ham"]
        src_classes = {"räucherschinken": 0, "ja": 1, "ich": 2, "esse": 3, "gern": 4}
        trg_classes = {"ham": 0, "smoked": 1, "i": 3, "love": 4, "to": 2, "eat": 4}
        corpus = [AlignedSent(trg_sentence, src_sentence)]
        alignment_info = AlignmentInfo(
            (0, 1, 4, 0, 2, 5, 5),
            [None] + src_sentence,
            ["UNUSED"] + trg_sentence,
            [[3], [1], [4], [], [2], [5, 6]],
        )

        head_vacancy_table = defaultdict(
            lambda: defaultdict(lambda: defaultdict(float))
        )
        head_vacancy_table[1 - 0][6][3] = 0.97  # ich -> i
        head_vacancy_table[3 - 0][5][4] = 0.97  # esse -> eat
        head_vacancy_table[1 - 2][4][4] = 0.97  # gern -> love
        head_vacancy_table[2 - 0][2][1] = 0.97  # räucherschinken -> smoked

        non_head_vacancy_table = defaultdict(
            lambda: defaultdict(lambda: defaultdict(float))
        )
        non_head_vacancy_table[1 - 0][1][0] = 0.96  # räucherschinken -> ham

        translation_table = defaultdict(lambda: defaultdict(float))
        translation_table["i"]["ich"] = 0.98
        translation_table["love"]["gern"] = 0.98
        translation_table["to"][None] = 0.98
        translation_table["eat"]["esse"] = 0.98
        translation_table["smoked"]["räucherschinken"] = 0.98
        translation_table["ham"]["räucherschinken"] = 0.98

        fertility_table = defaultdict(lambda: defaultdict(float))
        fertility_table[1]["ich"] = 0.99
        fertility_table[1]["esse"] = 0.99
        fertility_table[0]["ja"] = 0.99
        fertility_table[1]["gern"] = 0.99
        fertility_table[2]["räucherschinken"] = 0.999
        fertility_table[1][None] = 0.99

        probabilities = {
            "p1": 0.167,
            "translation_table": translation_table,
            "fertility_table": fertility_table,
            "head_vacancy_table": head_vacancy_table,
            "non_head_vacancy_table": non_head_vacancy_table,
            "head_distortion_table": None,
            "non_head_distortion_table": None,
            "alignment_table": None,
        }

        model5 = IBMModel5(corpus, 0, src_classes, trg_classes, probabilities)

        # act
        probability = model5.prob_t_a_given_s(alignment_info)

        # assert
        null_generation = 5 * pow(0.167, 1) * pow(0.833, 4)
        fertility = 1 * 0.99 * 1 * 0.99 * 1 * 0.99 * 1 * 0.99 * 2 * 0.999
        lexical_translation = 0.98 * 0.98 * 0.98 * 0.98 * 0.98 * 0.98
        vacancy = 0.97 * 0.97 * 1 * 0.97 * 0.97 * 0.96
        expected_probability = (
            null_generation * fertility * lexical_translation * vacancy
        )
        self.assertEqual(round(probability, 4), round(expected_probability, 4))

    def test_prune(self):
        # arrange
        alignment_infos = [
            AlignmentInfo((1, 1), None, None, None),
            AlignmentInfo((1, 2), None, None, None),
            AlignmentInfo((2, 1), None, None, None),
            AlignmentInfo((2, 2), None, None, None),
            AlignmentInfo((0, 0), None, None, None),
        ]
        min_factor = IBMModel5.MIN_SCORE_FACTOR
        best_score = 0.9
        scores = {
            (1, 1): min(min_factor * 1.5, 1) * best_score,  # above threshold
            (1, 2): best_score,
            (2, 1): min_factor * best_score,  # at threshold
            (2, 2): min_factor * best_score * 0.5,  # low score
            (0, 0): min(min_factor * 1.1, 1) * 1.2,  # above threshold
        }
        corpus = [AlignedSent(["a"], ["b"])]
        original_prob_function = IBMModel4.model4_prob_t_a_given_s
        # mock static method
        IBMModel4.model4_prob_t_a_given_s = staticmethod(
            lambda a, model: scores[a.alignment]
        )
        model5 = IBMModel5(corpus, 0, None, None)

        # act
        pruned_alignments = model5.prune(alignment_infos)

        # assert
        self.assertEqual(len(pruned_alignments), 3)

        # restore static method
        IBMModel4.model4_prob_t_a_given_s = original_prob_function