Datasets:

Modalities:
Audio
Text
ArXiv:
Libraries:
Datasets
License:
polinaeterna HF staff commited on
Commit
58b647c
1 Parent(s): 3476ae7

add dataset metadate

Browse files
Files changed (1) hide show
  1. voxpopuli.py +61 -14
voxpopuli.py CHANGED
@@ -7,6 +7,39 @@ from tqdm.auto import tqdm
7
  import datasets
8
 
9
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
10
  _LANGUAGES = sorted(
11
  [
12
  "en", "de", "fr", "es", "pl", "it", "ro", "hu", "cs", "nl", "fi", "hr",
@@ -15,6 +48,14 @@ _LANGUAGES = sorted(
15
  )
16
  _LANGUAGES_V2 = [f"{x}_v2" for x in _LANGUAGES]
17
 
 
 
 
 
 
 
 
 
18
  _YEARS = list(range(2009, 2020 + 1))
19
 
20
  # unnecessary
@@ -22,7 +63,7 @@ _CONFIG_TO_LANGS = {
22
  "400k": _LANGUAGES,
23
  "100k": _LANGUAGES,
24
  "10k": _LANGUAGES,
25
- "asr": ["original"],
26
  }
27
 
28
  _CONFIG_TO_YEARS = {
@@ -43,7 +84,11 @@ _BASE_URL = "https://dl.fbaipublicfiles.com/voxpopuli/"
43
 
44
  _DATA_URL = _BASE_URL + "audios/{lang}_{year}.tar"
45
 
46
- _META_URL = _BASE_URL + "annotations/unlabelled_v2.tsv.gz"
 
 
 
 
47
 
48
 
49
  class VoxpopuliConfig(datasets.BuilderConfig):
@@ -57,24 +102,23 @@ class VoxpopuliConfig(datasets.BuilderConfig):
57
  """
58
  super().__init__(name=name, **kwargs)
59
  name = name.split("_")[0]
60
- self.languages = [name] if name in _LANGUAGES else _LANGUAGES
61
  self.years = _CONFIG_TO_YEARS[name]
62
 
63
 
64
  class Voxpopuli(datasets.GeneratorBasedBuilder):
65
- """The Voxpopuli dataset."""
66
 
67
- VERSION = datasets.Version("1.0.0") # TODO ??
68
  BUILDER_CONFIGS = [
69
  VoxpopuliConfig(
70
  name=name,
71
- # version=VERSION,
72
- description="", # TODO
73
  )
74
  for name in _LANGUAGES + _LANGUAGES_V2 + ["10k", "100k", "400k"]
75
  ]
76
  # DEFAULT_CONFIG_NAME = "400k"
77
- DEFAULT_WRITER_BATCH_SIZE = 256
78
 
79
  def _info(self):
80
  try:
@@ -98,14 +142,14 @@ class Voxpopuli(datasets.GeneratorBasedBuilder):
98
  }
99
  )
100
  return datasets.DatasetInfo(
101
- # description=_DESCRIPTION,
102
  features=features,
103
- # homepage=_HOMEPAGE,
104
- # license=_LICENSE,
105
- # citation=_CITATION,
106
  )
107
 
108
- def _read_metadata(self, metadata_path):
109
  # from https://github.com/facebookresearch/voxpopuli/blob/main/voxpopuli/get_unlabelled_data.py#L34
110
  def predicate(id_):
111
  is_plenary = id_.find("PLENARY") > -1
@@ -133,8 +177,11 @@ class Voxpopuli(datasets.GeneratorBasedBuilder):
133
 
134
  return metadata
135
 
 
 
 
136
  def _split_generators(self, dl_manager):
137
- metadata_path = dl_manager.download_and_extract(_META_URL)
138
 
139
  urls = [_DATA_URL.format(lang=language, year=year) for language in self.config.languages for year in self.config.years]
140
  dl_manager.download_config.num_proc = len(urls)
 
7
  import datasets
8
 
9
 
10
+ _DESCRIPTION = """
11
+ A large-scale multilingual speech corpus for representation learning, semi-supervised learning and interpretation.
12
+ """
13
+
14
+ _CITATION = """
15
+ @inproceedings{wang-etal-2021-voxpopuli,
16
+ title = "{V}ox{P}opuli: A Large-Scale Multilingual Speech Corpus for Representation Learning,
17
+ Semi-Supervised Learning and Interpretation",
18
+ author = "Wang, Changhan and
19
+ Riviere, Morgane and
20
+ Lee, Ann and
21
+ Wu, Anne and
22
+ Talnikar, Chaitanya and
23
+ Haziza, Daniel and
24
+ Williamson, Mary and
25
+ Pino, Juan and
26
+ Dupoux, Emmanuel",
27
+ booktitle = "Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics
28
+ and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)",
29
+ month = aug,
30
+ year = "2021",
31
+ publisher = "Association for Computational Linguistics",
32
+ url = "https://aclanthology.org/2021.acl-long.80",
33
+ doi = "10.18653/v1/2021.acl-long.80",
34
+ pages = "993--1003",
35
+ }
36
+ """
37
+
38
+ _HOMEPAGE = "https://github.com/facebookresearch/voxpopuli"
39
+
40
+ _LICENSE = "CC0, also see https://www.europarl.europa.eu/legal-notice/en/"
41
+
42
+
43
  _LANGUAGES = sorted(
44
  [
45
  "en", "de", "fr", "es", "pl", "it", "ro", "hu", "cs", "nl", "fi", "hr",
 
48
  )
49
  _LANGUAGES_V2 = [f"{x}_v2" for x in _LANGUAGES]
50
 
51
+ _ASR_LANGUAGES = [
52
+ "en", "de", "fr", "es", "pl", "it", "ro", "hu", "cs", "nl", "fi", "hr",
53
+ "sk", "sl", "et", "lt"
54
+ ]
55
+ _ASR_ACCENTED_LANGUAGES = [
56
+ "en_accented"
57
+ ]
58
+
59
  _YEARS = list(range(2009, 2020 + 1))
60
 
61
  # unnecessary
 
63
  "400k": _LANGUAGES,
64
  "100k": _LANGUAGES,
65
  "10k": _LANGUAGES,
66
+ "asr": _ASR_LANGUAGES, # + _ASR_ACCENTED_LANGUAGES
67
  }
68
 
69
  _CONFIG_TO_YEARS = {
 
84
 
85
  _DATA_URL = _BASE_URL + "audios/{lang}_{year}.tar"
86
 
87
+ _ASR_DATA_URL = _BASE_URL + "audios/original_{year}.tar"
88
+
89
+ _UNLABELLED_META_URL = _BASE_URL + "annotations/unlabelled_v2.tsv.gz"
90
+
91
+ _ASR_META_URL = _BASE_URL + "annotations/asr/asr_{lang}.tsv.gz"
92
 
93
 
94
  class VoxpopuliConfig(datasets.BuilderConfig):
 
102
  """
103
  super().__init__(name=name, **kwargs)
104
  name = name.split("_")[0]
105
+ self.languages = [name] if name in _LANGUAGES else _CONFIG_TO_LANGS[name]
106
  self.years = _CONFIG_TO_YEARS[name]
107
 
108
 
109
  class Voxpopuli(datasets.GeneratorBasedBuilder):
110
+ """The VoxPopuli dataset."""
111
 
112
+ VERSION = datasets.Version("1.3.0") # not sure
113
  BUILDER_CONFIGS = [
114
  VoxpopuliConfig(
115
  name=name,
116
+ version=datasets.Version("1.3.0"),
 
117
  )
118
  for name in _LANGUAGES + _LANGUAGES_V2 + ["10k", "100k", "400k"]
119
  ]
120
  # DEFAULT_CONFIG_NAME = "400k"
121
+ DEFAULT_WRITER_BATCH_SIZE = 256 # SET THIS TO A LOWER VALUE IF IT USES TOO MUCH RAM SPACE
122
 
123
  def _info(self):
124
  try:
 
142
  }
143
  )
144
  return datasets.DatasetInfo(
145
+ description=_DESCRIPTION,
146
  features=features,
147
+ homepage=_HOMEPAGE,
148
+ license=_LICENSE,
149
+ citation=_CITATION,
150
  )
151
 
152
+ def _read_metadata_unlabelled(self, metadata_path):
153
  # from https://github.com/facebookresearch/voxpopuli/blob/main/voxpopuli/get_unlabelled_data.py#L34
154
  def predicate(id_):
155
  is_plenary = id_.find("PLENARY") > -1
 
177
 
178
  return metadata
179
 
180
+ def _read_metadata_asr(self, metadata_paths):
181
+ pass
182
+
183
  def _split_generators(self, dl_manager):
184
+ metadata_path = dl_manager.download_and_extract(_UNLABELLED_META_URL)
185
 
186
  urls = [_DATA_URL.format(lang=language, year=year) for language in self.config.languages for year in self.config.years]
187
  dl_manager.download_config.num_proc = len(urls)