jimregan commited on
Commit
d8100c5
1 Parent(s): 9c88183

add script

Browse files
Files changed (1) hide show
  1. clarinpl_studio.py +161 -0
clarinpl_studio.py ADDED
@@ -0,0 +1,161 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # coding=utf-8
2
+ # Copyright 2021 The TensorFlow Datasets Authors and the HuggingFace Datasets Authors.
3
+ # Copyright 2021 Jim O'Regan
4
+ #
5
+ # Licensed under the Apache License, Version 2.0 (the "License");
6
+ # you may not use this file except in compliance with the License.
7
+ # You may obtain a copy of the License at
8
+ #
9
+ # http://www.apache.org/licenses/LICENSE-2.0
10
+ #
11
+ # Unless required by applicable law or agreed to in writing, software
12
+ # distributed under the License is distributed on an "AS IS" BASIS,
13
+ # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14
+ # See the License for the specific language governing permissions and
15
+ # limitations under the License.
16
+
17
+ # Lint as: python3
18
+ """ClarinPL Studio automatic speech recognition dataset."""
19
+
20
+ import os
21
+ import glob
22
+
23
+ import datasets
24
+
25
+
26
+ _CITATION = """\
27
+ @article{korvzinek2017polish,
28
+ title={Polish read speech corpus for speech tools and services},
29
+ author={Kor{\v{z}}inek, Danijel and Marasek, Krzysztof and Brocki, {\L}ukasz and Wo{\l}k, Krzysztof},
30
+ journal={arXiv preprint arXiv:1706.00245},
31
+ year={2017}
32
+ }
33
+ """
34
+
35
+ _DESCRIPTION = """\
36
+ The corpus consists of 317 speakers recorded in 554
37
+ sessions, where each session consists of 20 read sentences and 10 phonetically rich words. The size of
38
+ the audio portion of the corpus amounts to around 56 hours, with transcriptions containing 356674 words
39
+ from a vocabulary of size 46361.
40
+
41
+ Note that in order to limit the required storage for preparing this dataset, the audio
42
+ is stored in the .wav format and is not converted to a float32 array. To convert the audio
43
+ file to a float32 array, please make use of the `.map()` function as follows:
44
+
45
+ ```python
46
+ import soundfile as sf
47
+
48
+ def map_to_array(batch):
49
+ speech_array, _ = sf.read(batch["file"])
50
+ batch["speech"] = speech_array
51
+ return batch
52
+
53
+ dataset = dataset.map(map_to_array, remove_columns=["file"])
54
+ ```
55
+ """
56
+
57
+ _URL = "https://mowa.clarin-pl.eu/"
58
+ _DS_URL = "http://mowa.clarin-pl.eu/korpusy/audio.tar.gz"
59
+ _TRAIN_URL = "https://raw.githubusercontent.com/danijel3/ClarinStudioKaldi/master/local_clarin/train.sessions"
60
+ _TEST_URL = "https://raw.githubusercontent.com/danijel3/ClarinStudioKaldi/master/local_clarin/test.sessions"
61
+ _VALID_URL = "https://raw.githubusercontent.com/danijel3/ClarinStudioKaldi/master/local_clarin/dev.sessions"
62
+
63
+ class ClarinPLStudioASRConfig(datasets.BuilderConfig):
64
+ """BuilderConfig for ClarinPLStudioASR."""
65
+
66
+ def __init__(self, **kwargs):
67
+ """
68
+ Args:
69
+ data_dir: `string`, the path to the folder containing the files in the
70
+ downloaded .tar
71
+ citation: `string`, citation for the data set
72
+ url: `string`, url for information about the data set
73
+ **kwargs: keyword arguments forwarded to super.
74
+ """
75
+ super(ClarinPLStudioASRConfig, self).__init__(version=datasets.Version("2.1.0", ""), **kwargs)
76
+
77
+
78
+ class ClarinPLStudio(datasets.GeneratorBasedBuilder):
79
+ """ClarinPL Studio dataset."""
80
+
81
+ BUILDER_CONFIGS = [
82
+ ClarinPLStudioASRConfig(name="clean", description="'Clean' speech."),
83
+ ]
84
+
85
+ def _info(self):
86
+ return datasets.DatasetInfo(
87
+ description=_DESCRIPTION,
88
+ features=datasets.Features(
89
+ {
90
+ "file": datasets.Value("string"),
91
+ "text": datasets.Value("string"),
92
+ "speaker_id": datasets.Value("string"),
93
+ "id": datasets.Value("string"),
94
+ }
95
+ ),
96
+ supervised_keys=("file", "text"),
97
+ homepage=_URL,
98
+ citation=_CITATION,
99
+ )
100
+
101
+ def _split_generators(self, dl_manager):
102
+ def get_sessions(path):
103
+ sessions = []
104
+ with open(path, 'r') as f:
105
+ for line in f:
106
+ sessions.append(line.strip())
107
+ return sessions
108
+ archive_path = dl_manager.download_and_extract(_DS_URL)
109
+ train_sessions_path = dl_manager.download(_TRAIN_URL)
110
+ test_sessions_path = dl_manager.download(_TEST_URL)
111
+ valid_sessions_path = dl_manager.download(_VALID_URL)
112
+
113
+ train_sessions = get_sessions(train_sessions_path)
114
+ test_sessions = get_sessions(test_sessions_path)
115
+ valid_sessions = get_sessions(valid_sessions_path)
116
+
117
+ archive_path = os.path.join(archive_path, "audio")
118
+ return [
119
+ datasets.SplitGenerator(name="train", gen_kwargs={
120
+ "archive_path": archive_path,
121
+ "sessions": train_sessions
122
+ }),
123
+ datasets.SplitGenerator(name="test", gen_kwargs={
124
+ "archive_path": archive_path,
125
+ "sessions": test_sessions
126
+ }),
127
+ datasets.SplitGenerator(name="valid", gen_kwargs={
128
+ "archive_path": archive_path,
129
+ "sessions": valid_sessions
130
+ }),
131
+ ]
132
+
133
+ def _generate_examples(self, archive_path, sessions):
134
+ """Generate examples from a ClarinPL Studio archive_path."""
135
+ def get_single_line(path):
136
+ lines = []
137
+ with open(path, 'r', encoding="utf-8") as f:
138
+ for line in f:
139
+ line = line.strip()
140
+ lines.append(line)
141
+ assert(len(lines) == 1)
142
+ return lines[0]
143
+ for session in sessions:
144
+ session_path = os.path.join(archive_path, session)
145
+ speaker = get_single_line(os.path.join(session_path, "spk.txt"))
146
+ text_glob = os.path.join(session_path, "*.txt")
147
+ for text_file in sorted(glob.glob(text_glob)):
148
+ if text_file.endswith("spk.txt"):
149
+ continue
150
+ basename = os.path.basename(text_file)
151
+ basename = basename.replace('.txt', '')
152
+ key = f'{session}_{basename}'
153
+ text = get_single_line(text_file)
154
+ audio = text_file.replace('.txt', '.wav')
155
+ example = {
156
+ "id": key,
157
+ "speaker_id": speaker,
158
+ "file": audio,
159
+ "text": text,
160
+ }
161
+ yield key, example