-
Notifications
You must be signed in to change notification settings - Fork 388
/
wer.py
144 lines (123 loc) · 5.14 KB
/
wer.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
# Copyright The PyTorch 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 Any, Callable, List, Optional, Union
import torch
from deprecate import deprecated, void
from torch import Tensor, tensor
from torchmetrics.functional.text.wer import _wer_compute, _wer_update
from torchmetrics.metric import Metric
from torchmetrics.utilities import _future_warning
class WordErrorRate(Metric):
r"""
Word error rate (WER_) is a common metric of the performance of an automatic speech recognition system.
This value indicates the percentage of words that were incorrectly predicted.
The lower the value, the better the performance of the ASR system with a WER of 0 being a perfect score.
Word error rate can then be computed as:
.. math::
WER = \frac{S + D + I}{N} = \frac{S + D + I}{S + D + C}
where:
- S is the number of substitutions,
- D is the number of deletions,
- I is the number of insertions,
- C is the number of correct words,
- N is the number of words in the reference (N=S+D+C).
Compute WER score of transcribed segments against references.
Args:
compute_on_step:
Forward only calls ``update()`` and return None if this is set to False.
dist_sync_on_step:
Synchronize metric state across processes at each ``forward()``
before returning the value at the step.
process_group:
Specify the process group on which synchronization is called.
dist_sync_fn:
Callback that performs the allgather operation on the metric state. When ``None``, DDP
will be used to perform the allgather
Returns:
Word error rate score
Examples:
>>> preds = ["this is the prediction", "there is an other sample"]
>>> target = ["this is the reference", "there is another one"]
>>> metric = WordErrorRate()
>>> metric(preds, target)
tensor(0.5000)
"""
is_differentiable = False
higher_is_better = False
error: Tensor
total: Tensor
def __init__(
self,
compute_on_step: bool = True,
dist_sync_on_step: bool = False,
process_group: Optional[Any] = None,
dist_sync_fn: Callable = None,
):
super().__init__(
compute_on_step=compute_on_step,
dist_sync_on_step=dist_sync_on_step,
process_group=process_group,
dist_sync_fn=dist_sync_fn,
)
self.add_state("errors", tensor(0, dtype=torch.float), dist_reduce_fx="sum")
self.add_state("total", tensor(0, dtype=torch.float), dist_reduce_fx="sum")
@deprecated(
args_mapping={"predictions": "preds", "references": "target"},
target=True,
deprecated_in="0.7",
remove_in="0.8",
stream=_future_warning,
)
def update(self, preds: Union[str, List[str]], target: Union[str, List[str]]) -> None: # type: ignore
"""Store references/predictions for computing Word Error Rate scores.
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
.. deprecated:: v0.7
Args:
predictions:
This argument is deprecated in favor of `preds` and will be removed in v0.8.
references:
This argument is deprecated in favor of `target` and will be removed in v0.8.
"""
errors, total = _wer_update(preds, target)
self.errors += errors
self.total += total
def compute(self) -> Tensor:
"""Calculate the word error rate.
Returns:
Word error rate score
"""
return _wer_compute(self.errors, self.total)
class WER(WordErrorRate):
r"""
Word error rate (WER_) is a common metric of the performance of an automatic speech recognition system.
.. deprecated:: v0.7
Use :class:`torchmetrics.WordErrorRate`. Will be removed in v0.8.
Examples:
>>> preds = ["this is the prediction", "there is an other sample"]
>>> target = ["this is the reference", "there is another one"]
>>> metric = WER()
>>> metric(preds, target)
tensor(0.5000)
"""
@deprecated(target=WordErrorRate, deprecated_in="0.7", remove_in="0.8", stream=_future_warning)
def __init__(
self,
compute_on_step: bool = True,
dist_sync_on_step: bool = False,
process_group: Optional[Any] = None,
dist_sync_fn: Callable = None,
) -> None:
void(compute_on_step, dist_sync_on_step, process_group, dist_sync_fn)