Upload biulders.py
Browse files- biulders.py +124 -0
biulders.py
ADDED
@@ -0,0 +1,124 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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 |
+
"""Leading and Trailing Silences Removed Large Nepali ASR Dataset"""
|
15 |
+
|
16 |
+
import os
|
17 |
+
import csv
|
18 |
+
|
19 |
+
import datasets
|
20 |
+
|
21 |
+
|
22 |
+
_CITATION = """\
|
23 |
+
@inproceedings{kjartansson-etal-sltu2018,
|
24 |
+
title = {{Crowd-Sourced Speech Corpora for Javanese, Sundanese, Sinhala, Nepali, and Bangladeshi Bengali}},
|
25 |
+
author = {Oddur Kjartansson and Supheakmungkol Sarin and Knot Pipatsrisawat and Martin Jansche and Linne Ha},
|
26 |
+
booktitle = {Proc. The 6th Intl. Workshop on Spoken Language Technologies for Under-Resourced Languages (SLTU)},
|
27 |
+
year = {2018},
|
28 |
+
address = {Gurugram, India},
|
29 |
+
month = aug,
|
30 |
+
pages = {52--55},
|
31 |
+
URL = {http://dx.doi.org/10.21437/SLTU.2018-11}
|
32 |
+
}
|
33 |
+
"""
|
34 |
+
|
35 |
+
_DESCRIPTION = """\
|
36 |
+
This data set contains transcribed audio data for Nepali. The data set consists of flac files, and a TSV file. The file utt_spk_text.tsv contains a FileID, anonymized UserID and the transcription of audio in the file.
|
37 |
+
The data set has been manually quality checked, but there might still be errors.
|
38 |
+
The audio files are sampled at rate of 16KHz, and leading and trailing silences are trimmed using torchaudio's voice activity detection.
|
39 |
+
"""
|
40 |
+
|
41 |
+
# Official homepage for the dataset
|
42 |
+
_HOMEPAGE = "https://www.openslr.org/54/"
|
43 |
+
|
44 |
+
# The licence for the dataset
|
45 |
+
_LICENSE = "license:cc-by-sa-4.0"
|
46 |
+
|
47 |
+
# TODO: Add link to the official dataset URLs here
|
48 |
+
# The HuggingFace Datasets library doesn't host the datasets but only points to the original files.
|
49 |
+
# This can be an arbitrary nested dict/list of URLs (see below in `_split_generators` method)
|
50 |
+
|
51 |
+
_URL = "https://huggingface.co/datasets/SumitMdhr/SANT-ASR/resolve/main/"
|
52 |
+
_URLS = {
|
53 |
+
"zipfile": _URL + "CLEAN_DATA.zip",
|
54 |
+
"index_file": _URL + "metedata1.csv",
|
55 |
+
}
|
56 |
+
|
57 |
+
|
58 |
+
# TODO: Name of the dataset usually match the script name with CamelCase instead of snake_case
|
59 |
+
class OpenslrNepaliAsrCleaned(datasets.GeneratorBasedBuilder):
|
60 |
+
"""End Silences Removed Large Nepali ASR Dataset"""
|
61 |
+
|
62 |
+
VERSION = datasets.Version("1.0.0")
|
63 |
+
# It's not mandatory to have a default configuration. Just use one if it make sense.
|
64 |
+
DEFAULT_CONFIG_NAME = "original"
|
65 |
+
|
66 |
+
def _info(self):
|
67 |
+
features = datasets.Features(
|
68 |
+
{
|
69 |
+
"utterance_id": datasets.Value("string"),
|
70 |
+
"speaker_id": datasets.Value("string"),
|
71 |
+
"utterance": datasets.Audio(),
|
72 |
+
"transcription": datasets.Value("string"),
|
73 |
+
"num_frames": datasets.Value("int32"),
|
74 |
+
}
|
75 |
+
)
|
76 |
+
return datasets.DatasetInfo(
|
77 |
+
description=_DESCRIPTION,
|
78 |
+
# Here we define them above because they are different between the two configurations
|
79 |
+
features=features,
|
80 |
+
# If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
|
81 |
+
# specify them. They'll be used if as_supervised=True in builder.as_dataset.
|
82 |
+
# supervised_keys=("sentence", "label"),
|
83 |
+
# Homepage of the dataset for documentation
|
84 |
+
homepage=_HOMEPAGE,
|
85 |
+
# License for the dataset if available
|
86 |
+
license=_LICENSE,
|
87 |
+
# Citation for the dataset
|
88 |
+
citation=_CITATION,
|
89 |
+
task_templates=[
|
90 |
+
datasets.tasks.AutomaticSpeechRecognition(
|
91 |
+
audio_column="utterance", transcription_column="transcription"
|
92 |
+
)
|
93 |
+
],
|
94 |
+
)
|
95 |
+
|
96 |
+
def _split_generators(self, dl_manager):
|
97 |
+
index_file = dl_manager.download(_URLS["index_file"])
|
98 |
+
zip_paths = dl_manager.download(_URLS["zipfiles"])
|
99 |
+
audio_paths = dl_manager.extract(zip_paths)
|
100 |
+
for path in zip_paths:
|
101 |
+
if os.path.exists(path):
|
102 |
+
os.remove(path)
|
103 |
+
return [
|
104 |
+
datasets.SplitGenerator(
|
105 |
+
name=datasets.Split.TRAIN,
|
106 |
+
gen_kwargs={
|
107 |
+
"index_file": index_file,
|
108 |
+
"audio_paths": audio_paths,
|
109 |
+
},
|
110 |
+
),
|
111 |
+
]
|
112 |
+
|
113 |
+
def _generate_examples(self, index_file, audio_paths):
|
114 |
+
with open(index_file, encoding="utf-8") as f:
|
115 |
+
reader = csv.DictReader(f, delimiter="\t")
|
116 |
+
for key, row in enumerate(reader):
|
117 |
+
path = os.path.join(audio_paths, "CLEAN_DATA", row["utterance_id"])
|
118 |
+
yield key, {
|
119 |
+
"utterance_id": row["utterance_id"],
|
120 |
+
"speaker_id": row["speaker_id"],
|
121 |
+
"utterance": path,
|
122 |
+
"transcription": row["transcription"],
|
123 |
+
"num_frames": int(row["num_frames"]),
|
124 |
+
}
|