litagin commited on
Commit
ecd0afe
·
verified ·
1 Parent(s): 77573cb

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +12 -5
README.md CHANGED
@@ -45,12 +45,20 @@ tags:
45
  - A Japanese speech dataset from Japanese visual novels (Galgames) intended for training ASR (Automatic Speech Recognition) models like Whisper.
46
  - This dataset is a derivative work of [OOPPEENN/Galgame_Dataset](https://huggingface.co/datasets/OOPPEENN/Galgame_Dataset) (thanks to the original authors!), and modified the original dataset for ASR purpose (see [Modifications](#modifications))
47
 
 
 
 
 
 
 
 
 
48
  ## Dataset Details
49
 
50
  - **Size**:
51
- - **3,746,321** audio files (all with transcriptions)
52
- - **5,355** total hours
53
- - 115 tar files totaling **100.16 GB**, with each tar file (except the last) containing 32,768 audio-text pairs (OGG and TXT files), approximately 897 MB per tar file
54
  - **Language**: Japanese
55
  - **Format**:
56
  - [**WebDataset**](https://github.com/webdataset/webdataset) format (see [Dataset Structure](#dataset-structure))
@@ -58,7 +66,7 @@ tags:
58
 
59
  ### Dataset Description
60
 
61
- - **Size**: 3,746,321 audio-text pairs, 5,355 hours, 100GB
62
  - **Language**: Japanese
63
  - **Format**: 16kHz, 16-bit, mono OGG
64
 
@@ -106,7 +114,6 @@ I modified the original dataset in the following ways:
106
  00000aa36e86ba49cb67fb886cce2c044c03dbb8ffddad4cb4e5f2da809e91ab.txt
107
  00000fe59140c18655921cd316f03ae7a81a0708a2d81a15d9b7ae866c459840.ogg
108
  00000fe59140c18655921cd316f03ae7a81a0708a2d81a15d9b7ae866c459840.txt
109
- 0000100db43be195b0dc774ab0397c3af594bd4c87683eeffa3df07c29f3eebc.ogg
110
  ...
111
  ```
112
 
 
45
  - A Japanese speech dataset from Japanese visual novels (Galgames) intended for training ASR (Automatic Speech Recognition) models like Whisper.
46
  - This dataset is a derivative work of [OOPPEENN/Galgame_Dataset](https://huggingface.co/datasets/OOPPEENN/Galgame_Dataset) (thanks to the original authors!), and modified the original dataset for ASR purpose (see [Modifications](#modifications))
47
 
48
+ ### Changelog
49
+
50
+ - 2024-10-12: Removed 190 audio-text pairs such that
51
+ - 🤗 Datasets cannot read for some reason (I don't know why...)
52
+ - Audio files longer than 30.0 seconds (for ASR trainig it's usual to filter such audio)
53
+
54
+ Resulting in 3,746,131 pairs and 5353.9 hours, and the number of files in each tar file may be smaller than 32768.
55
+
56
  ## Dataset Details
57
 
58
  - **Size**:
59
+ - **3,746,131** audio files (all with transcriptions)
60
+ - **5353.9** total hours
61
+ - 115 tar files totaling **100.16 GB**, with each tar file (except the last) containing about 32,768 audio-text pairs (OGG and TXT files), approximately 897 MB per tar file
62
  - **Language**: Japanese
63
  - **Format**:
64
  - [**WebDataset**](https://github.com/webdataset/webdataset) format (see [Dataset Structure](#dataset-structure))
 
66
 
67
  ### Dataset Description
68
 
69
+ - **Size**: 3,746,131 audio-text pairs, 5,353 hours, 100GB
70
  - **Language**: Japanese
71
  - **Format**: 16kHz, 16-bit, mono OGG
72
 
 
114
  00000aa36e86ba49cb67fb886cce2c044c03dbb8ffddad4cb4e5f2da809e91ab.txt
115
  00000fe59140c18655921cd316f03ae7a81a0708a2d81a15d9b7ae866c459840.ogg
116
  00000fe59140c18655921cd316f03ae7a81a0708a2d81a15d9b7ae866c459840.txt
 
117
  ...
118
  ```
119