yoom618 commited on
Commit
7d6fa61
1 Parent(s): ae1c8f4

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +39 -8
README.md CHANGED
@@ -7,6 +7,8 @@ language:
7
  size_categories:
8
  - 100K<n<1M
9
  ---
 
 
10
 
11
  ### Original Dataset can be found in <https://www.openslr.org/145/>
12
 
@@ -16,7 +18,7 @@ size_categories:
16
 
17
  ### How to Use
18
 
19
- - Almost the same with [Librispeech](https://huggingface.co/datasets/openslr/librispeech_asr) dataset module since i refered to the [source code](https://huggingface.co/datasets/openslr/librispeech_asr/blob/main/librispeech_asr.py).
20
 
21
  - three types of transcripts are given
22
  - `text_normalized` : the trascript from Librispeech ASR
@@ -31,17 +33,22 @@ import aiohttp
31
  import os
32
 
33
  # set huggingface cache directory for the extracted raw files and huggingface-cli token
34
- os.environ['HF_HOME'] = "/data/to/download"
35
- !export HF_HOME="/data/to/download"
 
36
 
37
 
38
  # download dataset
39
  # if already have librispeech_asr in the cache_dir it will use the same audio files.
40
- libripc = load_dataset("yoom618/librispeech_pc",
41
- "all", # all, clean, other
42
- cache_dir="/data/to/download",
43
  trust_remote_code=True,
44
- # storage_options={'client_kwargs': {'timeout': aiohttp.ClientTimeout(total=7200)}}, # add if you need to increase the timeout
 
 
 
 
45
 
46
 
47
  # check dataset info
@@ -55,6 +62,30 @@ pprint(libripc['train.clean.100'][0]['audio'])
55
  ```
56
 
57
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
58
 
59
  ### The number of samples in Librispeech-PC
60
 
@@ -62,7 +93,7 @@ pprint(libripc['train.clean.100'][0]['audio'])
62
  - `train.clean.100` : 26,041 ( 2,498 out of 28,539 were dropped )
63
  - `train.clean.360` : 95,404 ( 8,610 out of 104,014 were dropped )
64
  - `train.other.500` : 134,679 ( 14,009 out of 148,688 were dropped )
65
- - dev
66
  - `dev.clean` : 2,530 ( 173 out of 2,703 were dropped )
67
  - `dev.other` : 2,728 ( 136 out of 2,864 were dropped )
68
  - test
 
7
  size_categories:
8
  - 100K<n<1M
9
  ---
10
+ # Librispeech-PC (punctuation and capitalization restored)
11
+
12
 
13
  ### Original Dataset can be found in <https://www.openslr.org/145/>
14
 
 
18
 
19
  ### How to Use
20
 
21
+ - Almost the same with [Librispeech](https://huggingface.co/datasets/openslr/librispeech_asr) dataset module since i refered to its [source code](https://huggingface.co/datasets/openslr/librispeech_asr/blob/main/librispeech_asr.py).
22
 
23
  - three types of transcripts are given
24
  - `text_normalized` : the trascript from Librispeech ASR
 
33
  import os
34
 
35
  # set huggingface cache directory for the extracted raw files and huggingface-cli token
36
+ # if don't, you might download the raw tar.gz file in home cache dir even if you set `cache_dir` param
37
+ os.environ['HF_HOME'] = "/data_dir/to/download"
38
+ !export HF_HOME="/data_dir/to/download"
39
 
40
 
41
  # download dataset
42
  # if already have librispeech_asr in the cache_dir it will use the same audio files.
43
+ libripc = load_dataset("yoom618/librispeech_pc",
44
+ "all", # all, clean, other
45
+ cache_dir="/data_dir/to/download",
46
  trust_remote_code=True,
47
+ # storage_options={
48
+ # # add if you need to increase the timeout for openslr download
49
+ # 'client_kwargs': {'timeout': aiohttp.ClientTimeout(total=7200)}
50
+ # },
51
+ )
52
 
53
 
54
  # check dataset info
 
62
  ```
63
 
64
 
65
+ - Data Sample
66
+
67
+ ```raw
68
+ {'audio': {'array': array([ 7.01904297e-04, 7.32421875e-04, 7.32421875e-04, ...,
69
+ -2.74658203e-04, -1.83105469e-04, -3.05175781e-05]),
70
+ 'path': '/data_dir/to/download/downloads/extracted/.../374-180298-0000.flac',
71
+ 'sampling_rate': 16000},
72
+ 'chapter_id': 180298,
73
+ 'duration': 14.529999732971191,
74
+ 'file': '/data_dir/to/download/downloads/extracted/.../374-180298-0000.flac',
75
+ 'id': '374-180298-0000',
76
+ 'speaker_id': 374,
77
+ 'text': 'Chapter sixteen I might have told you of the beginning of this '
78
+ 'liaison in a few lines, but I wanted you to see every step by which '
79
+ 'we came, I to agree to whatever Marguerite wished,',
80
+ 'text_normalized': 'CHAPTER SIXTEEN I MIGHT HAVE TOLD YOU OF THE BEGINNING OF '
81
+ 'THIS LIAISON IN A FEW LINES BUT I WANTED YOU TO SEE EVERY '
82
+ 'STEP BY WHICH WE CAME I TO AGREE TO WHATEVER MARGUERITE '
83
+ 'WISHED',
84
+ 'text_raw': 'Chapter sixteen I might have told you of the beginning of this '
85
+ 'liaison in a few lines, but I wanted you to see every step by '
86
+ 'which we came, I to agree to whatever Marguerite wished,'}
87
+ ```
88
+
89
 
90
  ### The number of samples in Librispeech-PC
91
 
 
93
  - `train.clean.100` : 26,041 ( 2,498 out of 28,539 were dropped )
94
  - `train.clean.360` : 95,404 ( 8,610 out of 104,014 were dropped )
95
  - `train.other.500` : 134,679 ( 14,009 out of 148,688 were dropped )
96
+ - dev (validation)
97
  - `dev.clean` : 2,530 ( 173 out of 2,703 were dropped )
98
  - `dev.other` : 2,728 ( 136 out of 2,864 were dropped )
99
  - test