Add an initial version of the CVSS datasets loading script.
Browse files
cvss.py
ADDED
@@ -0,0 +1,148 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# Copyright 2022 The CVSS Dataset Authors and the dataset script contributor.
|
2 |
+
# All Rights Reserved.
|
3 |
+
#
|
4 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
5 |
+
# you may not use this file except in compliance with the License.
|
6 |
+
# You may obtain a copy of the License at
|
7 |
+
#
|
8 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
9 |
+
#
|
10 |
+
# Unless required by applicable law or agreed to in writing, software
|
11 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
12 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
13 |
+
# See the License for the specific language governing permissions and
|
14 |
+
# limitations under the License.
|
15 |
+
# ==============================================================================
|
16 |
+
"""CVSS speech-to-speech translation corpus."""
|
17 |
+
|
18 |
+
import os
|
19 |
+
import datasets
|
20 |
+
|
21 |
+
_HOMEPAGE = 'https://github.com/google-research-datasets/cvss'
|
22 |
+
|
23 |
+
_DESCRIPTION = """\
|
24 |
+
CVSS is a massively multilingual-to-English speech-to-speech translation corpus,
|
25 |
+
covering sentence-level parallel speech-to-speech translation pairs from 21
|
26 |
+
languages into English.
|
27 |
+
"""
|
28 |
+
|
29 |
+
_CITATION = """\
|
30 |
+
@inproceedings{jia2022cvss,
|
31 |
+
title={{CVSS} Corpus and Massively Multilingual Speech-to-Speech Translation},
|
32 |
+
author={Jia, Ye and Tadmor Ramanovich, Michelle and Wang, Quan and Zen, Heiga},
|
33 |
+
booktitle={Proceedings of Language Resources and Evaluation Conference (LREC)},
|
34 |
+
pages={6691--6703},
|
35 |
+
year={2022}
|
36 |
+
}
|
37 |
+
"""
|
38 |
+
|
39 |
+
_ROOT_URL = 'https://storage.googleapis.com/cvss'
|
40 |
+
|
41 |
+
_ALL_LANGUAGES = ('de', 'fr', 'es', 'ca', 'it', 'ru', 'zh', 'pt', 'fa', 'et',
|
42 |
+
'mn', 'nl', 'tr', 'ar', 'sv', 'lv', 'sl', 'ta', 'ja', 'id',
|
43 |
+
'cy')
|
44 |
+
|
45 |
+
|
46 |
+
def _get_download_urls(name='cvss_c', languages=_ALL_LANGUAGES, version='1.0'):
|
47 |
+
"""Gets URLs for downloading data.
|
48 |
+
|
49 |
+
Args:
|
50 |
+
name: 'cvss_c' or 'cvss_t'.
|
51 |
+
languages: An iterable of source languages.
|
52 |
+
version: Only '1.0' available.
|
53 |
+
|
54 |
+
Returns:
|
55 |
+
A list of URL strs.
|
56 |
+
"""
|
57 |
+
return [
|
58 |
+
f'{_ROOT_URL}/{name}_v{version}/{name}_{x}_en_v{version}.tar.gz'
|
59 |
+
for x in languages
|
60 |
+
]
|
61 |
+
|
62 |
+
|
63 |
+
class CVSSConfig(datasets.BuilderConfig):
|
64 |
+
"""BuilderConfig for CVSS."""
|
65 |
+
|
66 |
+
def __init__(self, name, languages='all', **kwargs):
|
67 |
+
"""BuilderConfig for CVSS.
|
68 |
+
|
69 |
+
Args:
|
70 |
+
name: 'cvss_c' or 'cvss_t'.
|
71 |
+
languages: A list of source languages.
|
72 |
+
**kwargs: keyword arguments forwarded to super.
|
73 |
+
"""
|
74 |
+
super().__init__(name=name, **kwargs)
|
75 |
+
if languages == 'all':
|
76 |
+
self.languages = _ALL_LANGUAGES
|
77 |
+
elif isinstance(languages, str):
|
78 |
+
self.languages = [languages]
|
79 |
+
else:
|
80 |
+
self.languages = languages
|
81 |
+
|
82 |
+
|
83 |
+
class CVSS(datasets.GeneratorBasedBuilder):
|
84 |
+
"""CVSS dataset. Version 1.0."""
|
85 |
+
|
86 |
+
BUILDER_CONFIG_CLASS = CVSSConfig
|
87 |
+
|
88 |
+
VERSION = '1.0.0'
|
89 |
+
|
90 |
+
DEFAULT_WRITER_BATCH_SIZE = 256
|
91 |
+
|
92 |
+
def _info(self):
|
93 |
+
return datasets.DatasetInfo(
|
94 |
+
description=_DESCRIPTION,
|
95 |
+
homepage=_HOMEPAGE,
|
96 |
+
citation=_CITATION,
|
97 |
+
features=datasets.Features({
|
98 |
+
'id': datasets.Value('string'),
|
99 |
+
'file': datasets.Value('string'),
|
100 |
+
'audio': datasets.Audio(sampling_rate=24_000),
|
101 |
+
'text': datasets.Value('string'),
|
102 |
+
}))
|
103 |
+
|
104 |
+
def _split_generators(self, dl_manager):
|
105 |
+
print(self.config)
|
106 |
+
downloaded_files = dl_manager.download_and_extract(
|
107 |
+
_get_download_urls(self.config.name, self.config.languages))
|
108 |
+
return [
|
109 |
+
datasets.SplitGenerator(
|
110 |
+
name=datasets.Split.TRAIN,
|
111 |
+
gen_kwargs={
|
112 |
+
'files': downloaded_files,
|
113 |
+
'split': 'train'
|
114 |
+
}),
|
115 |
+
datasets.SplitGenerator(
|
116 |
+
name=datasets.Split.VALIDATION,
|
117 |
+
gen_kwargs={
|
118 |
+
'files': downloaded_files,
|
119 |
+
'split': 'dev'
|
120 |
+
}),
|
121 |
+
datasets.SplitGenerator(
|
122 |
+
name=datasets.Split.TEST,
|
123 |
+
gen_kwargs={
|
124 |
+
'files': downloaded_files,
|
125 |
+
'split': 'test'
|
126 |
+
}),
|
127 |
+
]
|
128 |
+
|
129 |
+
def _generate_examples(self, files, split):
|
130 |
+
"""Generates examples for each SplitGenerator."""
|
131 |
+
for path in files:
|
132 |
+
with open(os.path.join(path, f'{split}.tsv'), 'r', encoding='utf-8') as f:
|
133 |
+
for line in f:
|
134 |
+
cols = line.rstrip().split('\t')
|
135 |
+
assert len(cols) == 2, cols
|
136 |
+
key, text = cols
|
137 |
+
|
138 |
+
audio_path = os.path.join(path, split, f'{key}.wav')
|
139 |
+
with open(audio_path, 'rb') as audio_f:
|
140 |
+
audio_bytes = audio_f.read()
|
141 |
+
|
142 |
+
audio = {'path': audio_path, 'bytes': audio_bytes}
|
143 |
+
yield key, {
|
144 |
+
'id': key,
|
145 |
+
'text': text,
|
146 |
+
'audio': audio,
|
147 |
+
'file': audio_path,
|
148 |
+
}
|