Datasets:
Tasks:
Automatic Speech Recognition
Formats:
parquet
Languages:
English
Size:
1M - 10M
ArXiv:
License:
Delete peoples_speech.py
Browse files- peoples_speech.py +0 -225
peoples_speech.py
DELETED
@@ -1,225 +0,0 @@
|
|
1 |
-
# Copyright 2020 The HuggingFace Datasets Authors and the current dataset script contributor.
|
2 |
-
#
|
3 |
-
# Licensed under the Apache License, Version 2.0 (the "License");
|
4 |
-
# you may not use this file except in compliance with the License.
|
5 |
-
# You may obtain a copy of the License at
|
6 |
-
#
|
7 |
-
# http://www.apache.org/licenses/LICENSE-2.0
|
8 |
-
#
|
9 |
-
# Unless required by applicable law or agreed to in writing, software
|
10 |
-
# distributed under the License is distributed on an "AS IS" BASIS,
|
11 |
-
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
12 |
-
# See the License for the specific language governing permissions and
|
13 |
-
# limitations under the License.
|
14 |
-
|
15 |
-
import json
|
16 |
-
import os
|
17 |
-
|
18 |
-
import datasets
|
19 |
-
from tqdm.auto import tqdm
|
20 |
-
|
21 |
-
|
22 |
-
# Find for instance the citation on arxiv or on the dataset repo/website
|
23 |
-
_CITATION = """\
|
24 |
-
@article{DBLP:journals/corr/abs-2111-09344,
|
25 |
-
author = {Daniel Galvez and
|
26 |
-
Greg Diamos and
|
27 |
-
Juan Ciro and
|
28 |
-
Juan Felipe Ceron and
|
29 |
-
Keith Achorn and
|
30 |
-
Anjali Gopi and
|
31 |
-
David Kanter and
|
32 |
-
Maximilian Lam and
|
33 |
-
Mark Mazumder and
|
34 |
-
Vijay Janapa Reddi},
|
35 |
-
title = {The People's Speech: A Large-Scale Diverse English Speech Recognition
|
36 |
-
Dataset for Commercial Usage},
|
37 |
-
journal = {CoRR},
|
38 |
-
volume = {abs/2111.09344},
|
39 |
-
year = {2021},
|
40 |
-
url = {https://arxiv.org/abs/2111.09344},
|
41 |
-
eprinttype = {arXiv},
|
42 |
-
eprint = {2111.09344},
|
43 |
-
timestamp = {Mon, 22 Nov 2021 16:44:07 +0100},
|
44 |
-
biburl = {https://dblp.org/rec/journals/corr/abs-2111-09344.bib},
|
45 |
-
bibsource = {dblp computer science bibliography, https://dblp.org}
|
46 |
-
}
|
47 |
-
"""
|
48 |
-
|
49 |
-
# You can copy an official description
|
50 |
-
_DESCRIPTION = """\
|
51 |
-
The People's Speech is a free-to-download 30,000-hour and growing supervised
|
52 |
-
conversational English speech recognition dataset licensed for academic and
|
53 |
-
commercial usage under CC-BY-SA (with a CC-BY subset).
|
54 |
-
"""
|
55 |
-
|
56 |
-
_HOMEPAGE = "https://mlcommons.org/en/peoples-speech/"
|
57 |
-
|
58 |
-
_LICENSE = [
|
59 |
-
"cc-by-2.0", "cc-by-2.5", "cc-by-3.0", "cc-by-4.0", "cc-by-sa-2.5",
|
60 |
-
"cc-by-sa-3.0", "cc-by-sa-4.0"
|
61 |
-
]
|
62 |
-
|
63 |
-
_BASE_URL = "https://huggingface.co/datasets/MLCommons/peoples_speech/resolve/main/"
|
64 |
-
|
65 |
-
# relative path to data inside dataset's repo
|
66 |
-
_DATA_URL = _BASE_URL + "{split}/{config}/{config}_{archive_id:06d}.tar"
|
67 |
-
|
68 |
-
# relative path to file containing number of audio archives inside dataset's repo
|
69 |
-
_N_SHARDS_URL = _BASE_URL + "n_shards.json"
|
70 |
-
|
71 |
-
# relative path to metadata inside dataset's repo
|
72 |
-
_MANIFEST_URL = _BASE_URL + "{split}/{config}.json"
|
73 |
-
|
74 |
-
|
75 |
-
class PeoplesSpeechConfig(datasets.BuilderConfig):
|
76 |
-
|
77 |
-
def __init__(self, *args, **kwargs):
|
78 |
-
super().__init__(*args, **kwargs)
|
79 |
-
|
80 |
-
|
81 |
-
class PeoplesSpeech(datasets.GeneratorBasedBuilder):
|
82 |
-
"""The People's Speech dataset."""
|
83 |
-
|
84 |
-
VERSION = datasets.Version("1.1.0")
|
85 |
-
BUILDER_CONFIGS = [
|
86 |
-
PeoplesSpeechConfig(name="microset", version=VERSION, description="Small subset of clean data for example pusposes."),
|
87 |
-
PeoplesSpeechConfig(name="clean", version=VERSION, description="Clean, CC-BY licensed subset."),
|
88 |
-
PeoplesSpeechConfig(name="dirty", version=VERSION, description="Dirty, CC-BY licensed subset."),
|
89 |
-
PeoplesSpeechConfig(name="clean_sa", version=VERSION, description="Clean, CC-BY-SA licensed subset."),
|
90 |
-
PeoplesSpeechConfig(name="dirty_sa", version=VERSION, description="Dirty, CC-BY-SA licensed subset."),
|
91 |
-
PeoplesSpeechConfig(name="test", version=VERSION, description="Only test data."),
|
92 |
-
PeoplesSpeechConfig(name="validation", version=VERSION, description="Only validation data."),
|
93 |
-
]
|
94 |
-
DEFAULT_CONFIG_NAME = "clean"
|
95 |
-
DEFAULT_WRITER_BATCH_SIZE = 512
|
96 |
-
|
97 |
-
def _info(self):
|
98 |
-
return datasets.DatasetInfo(
|
99 |
-
description=_DESCRIPTION,
|
100 |
-
features=datasets.Features(
|
101 |
-
{
|
102 |
-
"id": datasets.Value("string"),
|
103 |
-
"audio": datasets.Audio(),
|
104 |
-
"duration_ms": datasets.Value("int32"),
|
105 |
-
"text": datasets.Value("string"),
|
106 |
-
}
|
107 |
-
),
|
108 |
-
homepage=_HOMEPAGE,
|
109 |
-
license="/".join(_LICENSE), # license must be a string
|
110 |
-
citation=_CITATION,
|
111 |
-
)
|
112 |
-
|
113 |
-
def _split_generators(self, dl_manager):
|
114 |
-
|
115 |
-
if self.config.name == "microset":
|
116 |
-
# take only first data archive for demo purposes
|
117 |
-
url = [_DATA_URL.format(split="train", config="clean", archive_id=0)]
|
118 |
-
archive_path = dl_manager.download(url)
|
119 |
-
local_extracted_archive_path = dl_manager.extract(archive_path) if not dl_manager.is_streaming else [None]
|
120 |
-
manifest_url = _MANIFEST_URL.format(split="train", config="clean_000000") # train/clean_000000.json
|
121 |
-
manifest_path = dl_manager.download_and_extract(manifest_url)
|
122 |
-
|
123 |
-
return [
|
124 |
-
datasets.SplitGenerator(
|
125 |
-
name=datasets.Split.TRAIN,
|
126 |
-
gen_kwargs={
|
127 |
-
"local_extracted_archive_paths": local_extracted_archive_path,
|
128 |
-
# use iter_archive here to access the files in the TAR archives:
|
129 |
-
"archives": [dl_manager.iter_archive(path) for path in archive_path],
|
130 |
-
"manifest_path": manifest_path,
|
131 |
-
},
|
132 |
-
),
|
133 |
-
]
|
134 |
-
|
135 |
-
n_shards_path = dl_manager.download_and_extract(_N_SHARDS_URL)
|
136 |
-
with open(n_shards_path, encoding="utf-8") as f:
|
137 |
-
n_shards = json.load(f)
|
138 |
-
|
139 |
-
if self.config.name in ["validation", "test"]:
|
140 |
-
splits_to_configs = {self.config.name: self.config.name}
|
141 |
-
else:
|
142 |
-
splits_to_configs = {
|
143 |
-
"train": self.config.name,
|
144 |
-
"validation": "validation",
|
145 |
-
"test": "test"
|
146 |
-
}
|
147 |
-
|
148 |
-
audio_urls = {
|
149 |
-
split: [
|
150 |
-
_DATA_URL.format(split=split, config=config, archive_id=i) for i in range(n_shards[split][config])
|
151 |
-
] for split, config in splits_to_configs.items()
|
152 |
-
}
|
153 |
-
audio_archive_paths = dl_manager.download(audio_urls)
|
154 |
-
|
155 |
-
# In non-streaming mode, we extract the archives to have the data locally:
|
156 |
-
local_extracted_archive_paths = dl_manager.extract(audio_archive_paths) \
|
157 |
-
if not dl_manager.is_streaming else \
|
158 |
-
{split: [None] * len(audio_archive_paths[split]) for split in splits_to_configs}
|
159 |
-
|
160 |
-
manifest_urls = {
|
161 |
-
split: _MANIFEST_URL.format(split=split, config=config) for split, config in splits_to_configs.items()
|
162 |
-
}
|
163 |
-
manifest_paths = dl_manager.download_and_extract(manifest_urls)
|
164 |
-
|
165 |
-
# To access the audio data from the TAR archives using the download manager,
|
166 |
-
# we have to use the dl_manager.iter_archive method
|
167 |
-
#
|
168 |
-
# This is because dl_manager.download_and_extract
|
169 |
-
# doesn't work to stream TAR archives in streaming mode.
|
170 |
-
# (we have to stream the files of a TAR archive one by one)
|
171 |
-
#
|
172 |
-
# The iter_archive method returns an iterable of (path_within_archive, file_obj) for every
|
173 |
-
# file in a TAR archive.
|
174 |
-
splits_to_names = {
|
175 |
-
"train": datasets.Split.TRAIN,
|
176 |
-
"validation": datasets.Split.VALIDATION,
|
177 |
-
"test": datasets.Split.TEST,
|
178 |
-
}
|
179 |
-
split_generators = []
|
180 |
-
for split in splits_to_configs:
|
181 |
-
split_generators.append(
|
182 |
-
datasets.SplitGenerator(
|
183 |
-
name=splits_to_names[split],
|
184 |
-
gen_kwargs={
|
185 |
-
"local_extracted_archive_paths": local_extracted_archive_paths[split],
|
186 |
-
# use iter_archive here to access the files in the TAR archives:
|
187 |
-
"archives": [dl_manager.iter_archive(path) for path in audio_archive_paths[split]],
|
188 |
-
"manifest_path": manifest_paths[split],
|
189 |
-
}
|
190 |
-
)
|
191 |
-
)
|
192 |
-
|
193 |
-
return split_generators
|
194 |
-
|
195 |
-
def _generate_examples(self, local_extracted_archive_paths, archives, manifest_path):
|
196 |
-
meta = dict()
|
197 |
-
with open(manifest_path, "r", encoding="utf-8") as f:
|
198 |
-
for line in tqdm(f, desc="reading metadata file"):
|
199 |
-
sample_meta = json.loads(line)
|
200 |
-
_id = sample_meta["audio_document_id"]
|
201 |
-
texts = sample_meta["training_data"]["label"]
|
202 |
-
audio_filenames = sample_meta["training_data"]["name"]
|
203 |
-
durations = sample_meta["training_data"]["duration_ms"]
|
204 |
-
for audio_filename, text, duration in zip(audio_filenames, texts, durations):
|
205 |
-
audio_filename = audio_filename.lstrip("./")
|
206 |
-
meta[audio_filename] = {
|
207 |
-
"audio_document_id": _id,
|
208 |
-
"text": text,
|
209 |
-
"duration_ms": duration
|
210 |
-
}
|
211 |
-
|
212 |
-
for local_extracted_archive_path, archive in zip(local_extracted_archive_paths, archives):
|
213 |
-
# Here we iterate over all the files within the TAR archive:
|
214 |
-
for audio_filename, audio_file in archive:
|
215 |
-
audio_filename = audio_filename.lstrip("./")
|
216 |
-
# if an audio file exists locally (i.e. in default, non-streaming mode) set the full path to it
|
217 |
-
# joining path to directory that the archive was extracted to and audio filename.
|
218 |
-
path = os.path.join(local_extracted_archive_path, audio_filename) if local_extracted_archive_path \
|
219 |
-
else audio_filename
|
220 |
-
yield audio_filename, {
|
221 |
-
"id": audio_filename,
|
222 |
-
"audio": {"path": path, "bytes": audio_file.read()},
|
223 |
-
"text": meta[audio_filename]["text"],
|
224 |
-
"duration_ms": meta[audio_filename]["duration_ms"]
|
225 |
-
}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|