Datasets:

ArXiv:
License:
holylovenia commited on
Commit
f4cb165
·
verified ·
1 Parent(s): 5b0762a

Upload asr_malcsc.py with huggingface_hub

Browse files
Files changed (1) hide show
  1. asr_malcsc.py +195 -0
asr_malcsc.py ADDED
@@ -0,0 +1,195 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
+ import os
18
+ from pathlib import Path
19
+ from typing import Dict, List, Tuple
20
+
21
+ import datasets
22
+
23
+ from seacrowd.utils import schemas
24
+ from seacrowd.utils.configs import SEACrowdConfig
25
+ from seacrowd.utils.constants import Licenses, Tasks
26
+
27
+ # no bibtex citation
28
+ _CITATION = ""
29
+ _DATASETNAME = "asr_malcsc"
30
+ _DESCRIPTION = """\
31
+ This open-source dataset consists of 5 hours of transcribed Malay
32
+ conversational speech on certain topics, where ten conversations between five
33
+ pairs of speakers were contained.
34
+ """
35
+
36
+ _HOMEPAGE = "https://magichub.com/datasets/malay-conversational-speech-corpus/"
37
+ _LANGUAGES = ["zlm"]
38
+ _LICENSE = Licenses.CC_BY_NC_ND_4_0.value
39
+ _LOCAL = False
40
+ _URLS = {
41
+ _DATASETNAME: "https://magichub.com/df/df.php?file_name=Malay_Conversational_Speech_Corpus.zip",
42
+ }
43
+ _SUPPORTED_TASKS = [Tasks.SPEECH_RECOGNITION]
44
+
45
+ _SOURCE_VERSION = "1.0.0"
46
+ _SEACROWD_VERSION = "2024.06.20"
47
+
48
+
49
+ class ASRMalcscDataset(datasets.GeneratorBasedBuilder):
50
+ """ASR-Malcsc consists transcribed Malay conversational speech on certain topics"""
51
+
52
+ SOURCE_VERSION = datasets.Version(_SOURCE_VERSION)
53
+ SEACROWD_VERSION = datasets.Version(_SEACROWD_VERSION)
54
+
55
+ SEACROWD_SCHEMA_NAME = "sptext"
56
+
57
+ BUILDER_CONFIGS = [
58
+ SEACrowdConfig(
59
+ name=f"{_DATASETNAME}_source",
60
+ version=SOURCE_VERSION,
61
+ description=f"{_DATASETNAME} source schema",
62
+ schema="source",
63
+ subset_id=_DATASETNAME,
64
+ ),
65
+ SEACrowdConfig(
66
+ name=f"{_DATASETNAME}_seacrowd_{SEACROWD_SCHEMA_NAME}",
67
+ version=SEACROWD_VERSION,
68
+ description=f"{_DATASETNAME} SEACrowd schema",
69
+ schema=f"seacrowd_{SEACROWD_SCHEMA_NAME}",
70
+ subset_id=_DATASETNAME,
71
+ ),
72
+ ]
73
+
74
+ DEFAULT_CONFIG_NAME = f"{_DATASETNAME}_source"
75
+
76
+ def _info(self) -> datasets.DatasetInfo:
77
+
78
+ if self.config.schema == "source":
79
+ features = datasets.Features(
80
+ {
81
+ "id": datasets.Value("string"),
82
+ "channel": datasets.Value("string"),
83
+ "uttrans_id": datasets.Value("string"),
84
+ "speaker_id": datasets.Value("string"),
85
+ "topic": datasets.Value("string"),
86
+ "text": datasets.Value("string"),
87
+ "timestamp": datasets.Value("string"),
88
+ "path": datasets.Value("string"),
89
+ "audio": datasets.Audio(sampling_rate=16_000),
90
+ "speaker_gender": datasets.Value("string"),
91
+ "speaker_age": datasets.Value("int64"),
92
+ "speaker_region": datasets.Value("string"),
93
+ "speaker_device": datasets.Value("string"),
94
+ }
95
+ )
96
+
97
+ elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
98
+ features = schemas.speech_text_features
99
+
100
+ return datasets.DatasetInfo(
101
+ description=_DESCRIPTION,
102
+ features=features,
103
+ homepage=_HOMEPAGE,
104
+ license=_LICENSE,
105
+ citation=_CITATION,
106
+ )
107
+
108
+ def _split_generators(self, dl_manager: datasets.DownloadManager) -> List[datasets.SplitGenerator]:
109
+ """Returns SplitGenerators."""
110
+
111
+ data_paths = {
112
+ _DATASETNAME: Path(dl_manager.download_and_extract(_URLS[_DATASETNAME])),
113
+ }
114
+
115
+ return [
116
+ datasets.SplitGenerator(
117
+ name=datasets.Split.TRAIN,
118
+ gen_kwargs={
119
+ "filepath": data_paths[_DATASETNAME],
120
+ "split": "train",
121
+ },
122
+ )
123
+ ]
124
+
125
+ def _generate_examples(self, filepath: Path, split: str) -> Tuple[int, Dict]:
126
+ """Yields examples as (key, example) tuples."""
127
+
128
+ # read AUDIOINFO file
129
+ # columns: channel, uttrans_id, speaker_id, topic
130
+ audioinfo_filepath = os.path.join(filepath, "AUDIOINFO.txt")
131
+ with open(audioinfo_filepath, "r", encoding="utf-8") as audioinfo_file:
132
+ audioinfo_data = audioinfo_file.readlines()
133
+ audioinfo_data = audioinfo_data[1:] # remove header
134
+ audioinfo_data = [s.strip("\n").split("\t") for s in audioinfo_data]
135
+
136
+ # read SPKINFO file
137
+ # columns: channel, speaker_id, gender, age, region, device
138
+ spkinfo_filepath = os.path.join(filepath, "SPKINFO.txt")
139
+ with open(spkinfo_filepath, "r", encoding="utf-8") as spkinfo_file:
140
+ spkinfo_data = spkinfo_file.readlines()
141
+ spkinfo_data = spkinfo_data[1:] # remove header
142
+ spkinfo_data = [s.strip("\n").split("\t") for s in spkinfo_data]
143
+ for i, s in enumerate(spkinfo_data):
144
+ if s[2] == "M":
145
+ s[2] = "male"
146
+ elif s[2] == "F":
147
+ s[2] = "female"
148
+ else:
149
+ s[2] = None
150
+ # dictionary of metadata of each speaker
151
+ spkinfo_dict = {s[1]: {"speaker_gender": s[2], "speaker_age": int(s[3]), "speaker_region": s[4], "speaker_device": s[5]} for s in spkinfo_data}
152
+
153
+ num_sample = len(audioinfo_data)
154
+
155
+ for i in range(num_sample):
156
+ # wav file
157
+ wav_path = os.path.join(filepath, "WAV", audioinfo_data[i][1])
158
+ # transcription file
159
+ transcription_path = os.path.join(filepath, "TXT", audioinfo_data[i][1].replace("wav", "txt"))
160
+ with open(transcription_path, "r", encoding="utf-8") as transcription_file:
161
+ file_i = transcription_file.readlines()
162
+ # remove redundant speaker info from transcription file
163
+ file_i = [s.strip("\n").split("\t") for s in file_i]
164
+ transcription = [s[-1] for s in file_i]
165
+ timestamp = [s[0] for s in file_i]
166
+ text = " \n ".join(transcription)
167
+ timestamp_text = " \n ".join(timestamp)
168
+
169
+ if self.config.schema == "source":
170
+ example = {
171
+ "id": audioinfo_data[i][1].strip(".wav"),
172
+ "channel": audioinfo_data[i][0],
173
+ "uttrans_id": audioinfo_data[i][1],
174
+ "speaker_id": audioinfo_data[i][2],
175
+ "topic": audioinfo_data[i][3],
176
+ "text": text,
177
+ "timestamp": timestamp_text,
178
+ "path": wav_path,
179
+ "audio": wav_path,
180
+ "speaker_gender": spkinfo_dict[audioinfo_data[i][2]]["speaker_gender"],
181
+ "speaker_age": spkinfo_dict[audioinfo_data[i][2]]["speaker_age"],
182
+ "speaker_region": spkinfo_dict[audioinfo_data[i][2]]["speaker_region"],
183
+ "speaker_device": spkinfo_dict[audioinfo_data[i][2]]["speaker_device"],
184
+ }
185
+ elif self.config.schema == f"seacrowd_{self.SEACROWD_SCHEMA_NAME}":
186
+ example = {
187
+ "id": audioinfo_data[i][1].strip(".wav"),
188
+ "speaker_id": audioinfo_data[i][2],
189
+ "path": wav_path,
190
+ "audio": wav_path,
191
+ "text": text,
192
+ "metadata": {"speaker_age": spkinfo_dict[audioinfo_data[i][2]]["speaker_age"], "speaker_gender": spkinfo_dict[audioinfo_data[i][2]]["speaker_gender"]},
193
+ }
194
+
195
+ yield i, example