Datasets:

ArXiv:
License:
holylovenia commited on
Commit
ed939dc
1 Parent(s): 1552280

Upload ucla_phonetic.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. ucla_phonetic.py +158 -0
ucla_phonetic.py ADDED
@@ -0,0 +1,158 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2022 The HuggingFace Datasets Authors and the current dataset script contributor.
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
+ """
17
+ This dataset contains audio recordings and phonetic transcriptions of word utterances for various low-resource SEA languages.
18
+ Each language has a directory of text and audio files, with the latter forming one data subset.
19
+ The dataset is prepared from the online UCLA phonetic dataset, which contains 7000 utterances across 100 low-resource languages, phonetically aligned using various automatic approaches, and manually fixed for misalignments.
20
+ """
21
+ import os
22
+ from pathlib import Path
23
+ from typing import Dict, List, Tuple
24
+
25
+ import datasets
26
+
27
+ from seacrowd.utils import schemas
28
+ from seacrowd.utils.configs import SEACrowdConfig
29
+ from seacrowd.utils.constants import Licenses, Tasks
30
+
31
+ _CITATION = """\
32
+ @inproceedings{li2021multilingual,
33
+ title={Multilingual phonetic dataset for low resource speech recognition},
34
+ author={Li, Xinjian and Mortensen, David R and Metze, Florian and Black, Alan W},
35
+ booktitle={ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
36
+ pages={6958--6962},
37
+ year={2021},
38
+ organization={IEEE}
39
+ }
40
+ """
41
+
42
+ _DATASETNAME = "ucla_phonetic"
43
+
44
+ _DESCRIPTION = """\
45
+ This dataset contains audio recordings and phonetic transcriptions of word utterances for various low-resource SEA languages.
46
+ Each language has a directory of text and audio files, with the latter forming one data subset.
47
+ The dataset is prepared from the online UCLA phonetic dataset, which contains 7000 utterances across 100 low-resource languages, phonetically aligned using various automatic approaches, and manually fixed for misalignments.
48
+ """
49
+
50
+ _HOMEPAGE = "https://github.com/xinjli/ucla-phonetic-corpus"
51
+
52
+ _LANGUAGES = ["ace", "brv", "hil", "hni", "ilo", "khm", "mak", "mya", "pam"]
53
+
54
+ _LICENSE = Licenses.CC_BY_NC_SA_4_0.value
55
+
56
+ _LOCAL = False
57
+
58
+ _DATA_URL = "https://github.com/xinjli/ucla-phonetic-corpus/releases/download/v1.0/data.tar.gz"
59
+
60
+ _SUPPORTED_TASKS = [Tasks.SPEECH_RECOGNITION]
61
+
62
+ _SOURCE_VERSION = "1.0.0"
63
+ _SEACROWD_VERSION = "2024.06.20"
64
+
65
+
66
+ def seacrowd_config_constructor(lang, schema, version):
67
+ if lang not in _LANGUAGES:
68
+ raise ValueError(f"Invalid lang {lang}")
69
+
70
+ if schema not in ["source", "seacrowd_sptext"]:
71
+ raise ValueError(f"Invalid schema: {schema}")
72
+
73
+ return SEACrowdConfig(
74
+ name=f"ucla_phonetic_{lang}_{schema}",
75
+ version=datasets.Version(version),
76
+ description=f"UCLA Phonetic {schema} for {lang}",
77
+ schema=schema,
78
+ subset_id=f"{lang}_{schema}",
79
+ )
80
+
81
+
82
+ class UCLAPhoneticDataset(datasets.GeneratorBasedBuilder):
83
+ """This dataset contains audio recordings and phonetic transcriptions of word utterances for various low-resource SEA languages."""
84
+
85
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
86
+ SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
87
+
88
+ BUILDER_CONFIGS = (
89
+ [
90
+ SEACrowdConfig(
91
+ name="ucla_phonetic_source",
92
+ version=datasets.Version(_SOURCE_VERSION),
93
+ description="UCLA Phonetic source for ace",
94
+ schema="source",
95
+ subset_id="ace_source",
96
+ ),
97
+ SEACrowdConfig(
98
+ name="ucla_phonetic_seacrowd_sptext",
99
+ version=datasets.Version(_SOURCE_VERSION),
100
+ description="UCLA Phonetic seacrowd+sptext for ace",
101
+ schema="seacrowd_sptext",
102
+ subset_id="ace_seacrowd_sptext",
103
+ ),
104
+ ]
105
+ + [seacrowd_config_constructor(lang, "source", _SOURCE_VERSION) for lang in _LANGUAGES]
106
+ + [seacrowd_config_constructor(lang, "seacrowd_sptext", _SEACROWD_VERSION) for lang in _LANGUAGES]
107
+ )
108
+
109
+ DEFAULT_CONFIG_NAME = "ucla_phonetic_source"
110
+
111
+ def _info(self) -> datasets.DatasetInfo:
112
+
113
+ if self.config.schema == "source":
114
+ features = datasets.Features({"id": datasets.Value("string"), "text": datasets.Value("string"), "audio": datasets.Audio(sampling_rate=16_000)})
115
+ elif self.config.schema == "seacrowd_sptext":
116
+ features = schemas.speech_text_features
117
+
118
+ return datasets.DatasetInfo(
119
+ description=_DESCRIPTION,
120
+ features=features,
121
+ homepage=_HOMEPAGE,
122
+ license=_LICENSE,
123
+ citation=_CITATION,
124
+ )
125
+
126
+ def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
127
+ """Returns SplitGenerators."""
128
+
129
+ lang, schema = self.config.subset_id.split("_", maxsplit=1)
130
+ data_dir = dl_manager.download_and_extract(_DATA_URL)
131
+
132
+ return [
133
+ datasets.SplitGenerator(
134
+ name=datasets.Split.TRAIN,
135
+ # Whatever you put in gen_kwargs will be passed to _generate_examples
136
+ gen_kwargs={
137
+ "filepath": os.path.join(data_dir, "data", lang, "text.txt"),
138
+ "audiopath": Path(os.path.join(data_dir, "data", lang, "audio")),
139
+ },
140
+ )
141
+ ]
142
+
143
+ def _generate_examples(self, filepath: Path, audiopath: Path) -> Tuple[int, Dict]:
144
+
145
+ audiofiles = {}
146
+ for audiofile in audiopath.iterdir():
147
+ audio_idx = os.path.basename(audiofile).split(".")[0]
148
+ audiofiles[audio_idx] = audiofile
149
+
150
+ if self.config.schema == "source":
151
+ for line_idx, line in enumerate(open(filepath)):
152
+ audio_idx, text = line.strip().split(maxsplit=1)
153
+ yield line_idx, {"id": line_idx, "text": text, "audio": str(audiofiles[audio_idx])}
154
+
155
+ elif self.config.schema == "seacrowd_sptext":
156
+ for line_idx, line in enumerate(open(filepath)):
157
+ audio_idx, text = line.strip().split(maxsplit=1)
158
+ yield line_idx, {"id": line_idx, "path": str(audiofiles[audio_idx]), "audio": str(audiofiles[audio_idx]), "text": text, "speaker_id": None, "metadata": {"speaker_age": None, "speaker_gender": None}}