File size: 5,832 Bytes
a89d9fd
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#    http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from rapidfuzz.distance import Levenshtein
from difflib import SequenceMatcher

import numpy as np
import string


class RecMetric(object):
    def __init__(self,
                 main_indicator='acc',
                 is_filter=False,
                 ignore_space=True,
                 **kwargs):
        self.main_indicator = main_indicator
        self.is_filter = is_filter
        self.ignore_space = ignore_space
        self.eps = 1e-5
        self.reset()

    def _normalize_text(self, text):
        text = ''.join(
            filter(lambda x: x in (string.digits + string.ascii_letters), text))
        return text.lower()

    def __call__(self, pred_label, *args, **kwargs):
        preds, labels = pred_label
        correct_num = 0
        all_num = 0
        norm_edit_dis = 0.0
        for (pred, pred_conf), (target, _) in zip(preds, labels):
            if self.ignore_space:
                pred = pred.replace(" ", "")
                target = target.replace(" ", "")
            if self.is_filter:
                pred = self._normalize_text(pred)
                target = self._normalize_text(target)
            norm_edit_dis += Levenshtein.normalized_distance(pred, target)
            if pred == target:
                correct_num += 1
            all_num += 1
        self.correct_num += correct_num
        self.all_num += all_num
        self.norm_edit_dis += norm_edit_dis
        return {
            'acc': correct_num / (all_num + self.eps),
            'norm_edit_dis': 1 - norm_edit_dis / (all_num + self.eps)
        }

    def get_metric(self):
        """
        return metrics {
                 'acc': 0,
                 'norm_edit_dis': 0,
            }
        """
        acc = 1.0 * self.correct_num / (self.all_num + self.eps)
        norm_edit_dis = 1 - self.norm_edit_dis / (self.all_num + self.eps)
        self.reset()
        return {'acc': acc, 'norm_edit_dis': norm_edit_dis}

    def reset(self):
        self.correct_num = 0
        self.all_num = 0
        self.norm_edit_dis = 0


class CNTMetric(object):
    def __init__(self, main_indicator='acc', **kwargs):
        self.main_indicator = main_indicator
        self.eps = 1e-5
        self.reset()

    def __call__(self, pred_label, *args, **kwargs):
        preds, labels = pred_label
        correct_num = 0
        all_num = 0
        for pred, target in zip(preds, labels):
            if pred == target:
                correct_num += 1
            all_num += 1
        self.correct_num += correct_num
        self.all_num += all_num
        return {'acc': correct_num / (all_num + self.eps), }

    def get_metric(self):
        """
        return metrics {
                 'acc': 0,
            }
        """
        acc = 1.0 * self.correct_num / (self.all_num + self.eps)
        self.reset()
        return {'acc': acc}

    def reset(self):
        self.correct_num = 0
        self.all_num = 0


class CANMetric(object):
    def __init__(self, main_indicator='exp_rate', **kwargs):
        self.main_indicator = main_indicator
        self.word_right = []
        self.exp_right = []
        self.word_total_length = 0
        self.exp_total_num = 0
        self.word_rate = 0
        self.exp_rate = 0
        self.reset()
        self.epoch_reset()

    def __call__(self, preds, batch, **kwargs):
        for k, v in kwargs.items():
            epoch_reset = v
            if epoch_reset:
                self.epoch_reset()
        word_probs = preds
        word_label, word_label_mask = batch
        line_right = 0
        if word_probs is not None:
            word_pred = word_probs.argmax(2)
        word_pred = word_pred.cpu().detach().numpy()
        word_scores = [
            SequenceMatcher(
                None,
                s1[:int(np.sum(s3))],
                s2[:int(np.sum(s3))],
                autojunk=False).ratio() * (
                    len(s1[:int(np.sum(s3))]) + len(s2[:int(np.sum(s3))])) /
            len(s1[:int(np.sum(s3))]) / 2
            for s1, s2, s3 in zip(word_label, word_pred, word_label_mask)
        ]
        batch_size = len(word_scores)
        for i in range(batch_size):
            if word_scores[i] == 1:
                line_right += 1
        self.word_rate = np.mean(word_scores)  #float
        self.exp_rate = line_right / batch_size  #float
        exp_length, word_length = word_label.shape[:2]
        self.word_right.append(self.word_rate * word_length)
        self.exp_right.append(self.exp_rate * exp_length)
        self.word_total_length = self.word_total_length + word_length
        self.exp_total_num = self.exp_total_num + exp_length

    def get_metric(self):
        """
        return {
            'word_rate': 0,
            "exp_rate": 0,
        }
        """
        cur_word_rate = sum(self.word_right) / self.word_total_length
        cur_exp_rate = sum(self.exp_right) / self.exp_total_num
        self.reset()
        return {'word_rate': cur_word_rate, "exp_rate": cur_exp_rate}

    def reset(self):
        self.word_rate = 0
        self.exp_rate = 0

    def epoch_reset(self):
        self.word_right = []
        self.exp_right = []
        self.word_total_length = 0
        self.exp_total_num = 0