-
Notifications
You must be signed in to change notification settings - Fork 387
/
mer.py
88 lines (70 loc) · 2.96 KB
/
mer.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
# Copyright The Lightning team.
#
# 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 typing import List, Tuple, Union
import torch
from torch import Tensor, tensor
from torchmetrics.functional.text.helper import _edit_distance
def _mer_update(
preds: Union[str, List[str]],
target: Union[str, List[str]],
) -> Tuple[Tensor, Tensor]:
"""Update the mer score with the current set of references and predictions.
Args:
preds: Transcription(s) to score as a string or list of strings
target: Reference(s) for each speech input as a string or list of strings
Returns:
Number of edit operations to get from the reference to the prediction, summed over all samples
Number of words overall references
"""
if isinstance(preds, str):
preds = [preds]
if isinstance(target, str):
target = [target]
errors = tensor(0, dtype=torch.float)
total = tensor(0, dtype=torch.float)
for pred, tgt in zip(preds, target):
pred_tokens = pred.split()
tgt_tokens = tgt.split()
errors += _edit_distance(pred_tokens, tgt_tokens)
total += max(len(tgt_tokens), len(pred_tokens))
return errors, total
def _mer_compute(errors: Tensor, total: Tensor) -> Tensor:
"""Compute the match error rate.
Args:
errors: Number of edit operations to get from the reference to the prediction, summed over all samples
total: Number of words overall references
Returns:
Match error rate score
"""
return errors / total
def match_error_rate(preds: Union[str, List[str]], target: Union[str, List[str]]) -> Tensor:
"""Match error rate is a metric of the performance of an automatic speech recognition system.
This value indicates the percentage of words that were incorrectly predicted and inserted. The lower the value, the
better the performance of the ASR system with a MatchErrorRate of 0 being a perfect score.
Args:
preds: Transcription(s) to score as a string or list of strings
target: Reference(s) for each speech input as a string or list of strings
Returns:
Match error rate score
Examples:
>>> preds = ["this is the prediction", "there is an other sample"]
>>> target = ["this is the reference", "there is another one"]
>>> match_error_rate(preds=preds, target=target)
tensor(0.4444)
"""
errors, total = _mer_update(
preds,
target,
)
return _mer_compute(errors, total)