>>> wer_score = wer.compute(predictions=predictions, references=references)
>>> print(wer_score)
0.5
"""
class AddSpacesToPunctuation(tr.AbstractTransform):
def process_string(self, s: str):
return re.sub("(['.,:;?!&])", r" \1 ", s)
_transform = tr.Compose(
[
AddSpacesToPunctuation(),
tr.RemoveMultipleSpaces(),
tr.Strip(),
tr.SentencesToListOfWords(),
tr.RemoveEmptyStrings(),
]
)
class WER_punctuation(datasets.Metric):
def _info(self):
return datasets.MetricInfo(
description=_DESCRIPTION,
citation=_CITATION,
inputs_description=_KWARGS_DESCRIPTION,
features=datasets.Features(
{
"predictions": datasets.Value("string", id="sequence"),
def process_string(self, s: str):
return list(s)
def process_list(self, inp: List[str]):
chars = []
for sent_idx, sentence in enumerate(inp):
chars.extend(self.process_string(sentence))
if self.sentence_delimiter is not None and self.sentence_delimiter != "" and sent_idx < len(
inp) - 1:
chars.append(self.sentence_delimiter)
return chars
cer_transform = tr.Compose([
tr.RemoveMultipleSpaces(),
tr.Strip(),
SentencesToListOfCharacters(SENTENCE_DELIMITER)
])
else:
cer_transform = tr.Compose([
tr.RemoveMultipleSpaces(),
tr.Strip(),
tr.ReduceToSingleSentence(SENTENCE_DELIMITER),
tr.ReduceToListOfListOfChars(),
])
_CITATION = """\
@inproceedings{inproceedings,
author = {Morris, Andrew and Maier, Viktoria and Green, Phil},
year = {2004},
month = {01},
import jiwer.transforms as tr
__all__ = [
"wer_default",
"wer_contiguous",
"wer_standardize",
"wer_standardize_contiguous",
"cer_default_transform",
]
################################################################################
# implement transformations for WER (and accompanying measures)
wer_default = tr.Compose([
tr.RemoveMultipleSpaces(),
tr.Strip(),
tr.ReduceToListOfListOfWords(),
])
wer_contiguous = tr.Compose([
tr.RemoveMultipleSpaces(),
tr.Strip(),
tr.ReduceToSingleSentence(),
tr.ReduceToListOfListOfWords(),
])
wer_standardize = tr.Compose([
tr.ToLowerCase(),
tr.ExpandCommonEnglishContractions(),
tr.RemoveKaldiNonWords(),
tr.RemoveWhiteSpace(replace_by_space=True),
import Levenshtein
from typing import List, Mapping, Tuple, Union
import jiwer.transforms as tr
__all__ = ["wer", "mer", "wil", "wip", "compute_measures", "ops"]
################################################################################
# Implementation of the WER method, exposed publicly
_default_transform = tr.Compose(
[
tr.RemoveMultipleSpaces(),
tr.Strip(),
tr.SentencesToListOfWords(),
tr.RemoveEmptyStrings(),
]
)
_standardize_transform = tr.Compose(
[
tr.ToLowerCase(),
tr.ExpandCommonEnglishContractions(),
tr.RemoveKaldiNonWords(),
tr.RemoveWhiteSpace(replace_by_space=True),
]
)
def ops(
class SentencesToListOfCharacters(tr.AbstractTransform):
def process_string(self, s: str):
return list(s)
def process_list(self, inp: List[str]):
chars = []
for sentence in inp:
chars.extend(self.process_string(sentence))
return chars
cer_transform = tr.Compose([
tr.RemoveMultipleSpaces(),
tr.Strip(),
SentencesToListOfCharacters(),
])
_CITATION = """\
@inproceedings{inproceedings,
author = {Morris, Andrew and Maier, Viktoria and Green, Phil},
year = {2004},
month = {01},
pages = {},
title = {From WER and RIL to MER and WIL: improved evaluation measures for connected speech recognition.}
}
"""
_DESCRIPTION = """\
Character error rate (CER) is a common metric of the performance of an automatic speech recognition system.
