holylovenia commited on
Commit
d550b9a
1 Parent(s): 62f63b2

Upload bloom_speech.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. bloom_speech.py +172 -0
bloom_speech.py ADDED
@@ -0,0 +1,172 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """
2
+ SEA Crowd Data Loader for Bloom Speech.
3
+ """
4
+ from typing import Dict, List, Tuple
5
+
6
+ import datasets
7
+ from datasets.download.download_manager import DownloadManager
8
+
9
+ from seacrowd.utils import schemas
10
+ from seacrowd.utils.configs import SEACrowdConfig
11
+ from seacrowd.utils.constants import TASK_TO_SCHEMA, Licenses, Tasks
12
+
13
+ _CITATION = r"""
14
+ @inproceedings{leong-etal-2022-bloom,
15
+ title = "Bloom Library: Multimodal Datasets in 300+ Languages for a Variety of Downstream Tasks",
16
+ author = "Leong, Colin and
17
+ Nemecek, Joshua and
18
+ Mansdorfer, Jacob and
19
+ Filighera, Anna and
20
+ Owodunni, Abraham and
21
+ Whitenack, Daniel",
22
+ editor = "Goldberg, Yoav and
23
+ Kozareva, Zornitsa and
24
+ Zhang, Yue",
25
+ booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
26
+ month = dec,
27
+ year = "2022",
28
+ address = "Abu Dhabi, United Arab Emirates",
29
+ publisher = "Association for Computational Linguistics",
30
+ url = "https://aclanthology.org/2022.emnlp-main.590",
31
+ doi = "10.18653/v1/2022.emnlp-main.590",
32
+ pages = "8608--8621",
33
+ }
34
+ """
35
+
36
+ logger = datasets.logging.get_logger(__name__)
37
+
38
+ # this config is created for SEACrowd Dataloader
39
+ _LANG_CONFIG = {"bjn": "Banjar", "bzi": "Bisu", "ceb": "Cebuano", "ind": "Indonesian", "jra": "Jarai", "kqr": "Kimaragang", "mya": "Burmese", "tgl": "Tagalog"}
40
+
41
+ _LOCAL = False
42
+ _LANGUAGES = list(_LANG_CONFIG.keys())
43
+
44
+
45
+ _DATASETNAME = "bloom_speech"
46
+ _DESCRIPTION = r"""
47
+ This version of the Bloom Library data is developed specifically for the automatic speech recognition and speech-to-text tasks.
48
+ It includes data from 56 languages across 18 language families. 8 languages are spoken in Southeast Asia.
49
+ Before using this dataloader, please accept the acknowledgement at https://huggingface.co/datasets/sil-ai/bloom-speech and use huggingface-cli login for authentication.
50
+ """
51
+
52
+ _HOMEPAGE = "https://huggingface.co/datasets/sil-ai/bloom-speech"
53
+ _LICENSE = Licenses.CC.value
54
+
55
+ _URL = "https://huggingface.co/datasets/sil-ai/bloom-speech"
56
+ _HF_REMOTE_REF = "/".join(_URL.split("/")[-2:])
57
+
58
+ _SUPPORTED_TASKS = [Tasks.SPEECH_RECOGNITION]
59
+ _SOURCE_VERSION = "0.0.1"
60
+ _SEACROWD_VERSION = "2024.06.20"
61
+
62
+ CONFIG_SUFFIXES_FOR_TASK = [TASK_TO_SCHEMA.get(task).lower() for task in _SUPPORTED_TASKS]
63
+
64
+
65
+ def construct_configs_on_langs(languages: list = None) -> List[SEACrowdConfig]:
66
+ """
67
+ The function `construct_configs` constructs a list of SEACrowdConfig objects based on the provided
68
+ languages or a default language, and returns the list.
69
+
70
+ input:
71
+ languages (list, default None): The `languages` parameter is a list that specifies the languages for which the
72
+ configurations need to be constructed. If no languages are provided (value=None), the first value in language config
73
+ will be used.
74
+ output:
75
+ a list of `SEACrowdConfig` objects based on instantiated init variables
76
+ """
77
+
78
+ # set output var
79
+ config_list = []
80
+
81
+ # construct zipped arg for config instantiation
82
+ TASKS_AND_CONFIG_SUFFIX_PAIRS = list(zip(_SUPPORTED_TASKS, CONFIG_SUFFIXES_FOR_TASK))
83
+
84
+ # implement source schema
85
+ version, config_name_prefix = _SOURCE_VERSION, "source"
86
+ config_list += [
87
+ SEACrowdConfig(
88
+ name=f"{_DATASETNAME}_{_LANG}_{config_name_prefix}",
89
+ version=datasets.Version(version),
90
+ description=f"{_DATASETNAME} {config_name_prefix} schema for language code {_LANG}",
91
+ schema=f"{config_name_prefix}",
92
+ subset_id=_LANG,
93
+ )
94
+ for _LANG in languages
95
+ ]
96
+
97
+ # implement SEACrowd schema
98
+ version, config_name_prefix = _SEACROWD_VERSION, "seacrowd"
99
+ for task_obj, config_name_suffix in TASKS_AND_CONFIG_SUFFIX_PAIRS:
100
+ config_list += [
101
+ SEACrowdConfig(
102
+ name=f"{_DATASETNAME}_{_LANG}_{config_name_prefix}_{config_name_suffix}",
103
+ version=datasets.Version(version),
104
+ description=f"{_DATASETNAME} {config_name_prefix} schema for {task_obj.name} and language code {_LANG}",
105
+ schema=f"{config_name_prefix}_{config_name_suffix}",
106
+ subset_id=_LANG,
107
+ )
108
+ for _LANG in languages
109
+ ]
110
+ return config_list
111
+
112
+
113
+ class BloomSpeechDataset(datasets.GeneratorBasedBuilder):
114
+ """Bloom Speech dataset, subsetted from https://huggingface.co/datasets/sil-ai/bloom-speech"""
115
+
116
+ # get all schema w/o lang arg + get all schema w/ lang arg
117
+ BUILDER_CONFIGS = construct_configs_on_langs(_LANGUAGES)
118
+
119
+ def _info(self) -> datasets.DatasetInfo:
120
+ _config_schema_name = self.config.schema
121
+ logger.info(f"Received schema name: {self.config.schema}")
122
+ # source schema
123
+ if _config_schema_name == "source":
124
+ features = datasets.Features(
125
+ {
126
+ "file": datasets.Value("string"),
127
+ "audio": datasets.Audio(sampling_rate=16_000),
128
+ "text": datasets.Value("string"),
129
+ "book": datasets.Value("string"),
130
+ "instance": datasets.Value("string"),
131
+ "license": datasets.Value("string"),
132
+ "credits": datasets.Value("string"),
133
+ "original_lang_tag": datasets.Value("string"),
134
+ }
135
+ )
136
+
137
+ # speech-text schema
138
+ elif _config_schema_name == "seacrowd_sptext":
139
+ features = schemas.speech_text_features
140
+
141
+ else:
142
+ raise ValueError(f"Received unexpected config schema of {_config_schema_name}!")
143
+
144
+ return datasets.DatasetInfo(
145
+ description=_DESCRIPTION,
146
+ features=features,
147
+ homepage=_HOMEPAGE,
148
+ license=_LICENSE,
149
+ citation=_CITATION,
150
+ )
151
+
152
+ def _split_generators(self, dl_manager: DownloadManager) -> List[datasets.SplitGenerator]:
153
+ hf_dset_dict = datasets.load_dataset(_HF_REMOTE_REF, self.config.subset_id)
154
+
155
+ return [datasets.SplitGenerator(name=datasets.Split(dset_key), gen_kwargs={"hf_dset": dset}) for dset_key, dset in hf_dset_dict.items() if dset.num_rows > 0]
156
+
157
+ def _generate_examples(self, hf_dset) -> Tuple[int, Dict]:
158
+ _config_schema_name = self.config.schema
159
+
160
+ _idx = 0
161
+ for datapoints in hf_dset:
162
+ # since no _idx is available to be used, we're creating it manually for both schema
163
+ if _config_schema_name == "source":
164
+ yield _idx, {colname: datapoints[colname] for colname in self.info.features}
165
+
166
+ elif _config_schema_name == "seacrowd_sptext":
167
+ yield _idx, {"id": _idx, "path": datapoints["file"], "audio": datapoints["audio"], "text": datapoints["text"], "speaker_id": None, "metadata": {"speaker_age": None, "speaker_gender": None}}
168
+
169
+ else:
170
+ raise ValueError(f"Received unexpected config schema of {_config_schema_name}!")
171
+
172
+ _idx += 1