Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -98,6 +98,7 @@ I modified the original dataset in the following ways:
|
|
| 98 |
### Direct Use
|
| 99 |
|
| 100 |
- Fine-tuning ASR models like Whisper for the Japanese anime-like speech domain
|
|
|
|
| 101 |
- Training ASR models for the NSFW domain (aegi and chupa voices), which Whisper and other ASR models mostly cannot recognize
|
| 102 |
|
| 103 |
### Out-of-Scope Use
|
|
@@ -117,7 +118,7 @@ I modified the original dataset in the following ways:
|
|
| 117 |
...
|
| 118 |
```
|
| 119 |
|
| 120 |
-
- Except for the last tar file, each tar file contains 32768 audio-text pairs (OGG and TXT files), hence 65536 files in total.
|
| 121 |
|
| 122 |
- File names are randomly generated SHA-256 hashes, so the order of the files has no mean (e.g., the files coming from the same Galgame are not necessarily adjacent).
|
| 123 |
|
|
|
|
| 98 |
### Direct Use
|
| 99 |
|
| 100 |
- Fine-tuning ASR models like Whisper for the Japanese anime-like speech domain
|
| 101 |
+
- Benchmarking Japanese ASR models
|
| 102 |
- Training ASR models for the NSFW domain (aegi and chupa voices), which Whisper and other ASR models mostly cannot recognize
|
| 103 |
|
| 104 |
### Out-of-Scope Use
|
|
|
|
| 118 |
...
|
| 119 |
```
|
| 120 |
|
| 121 |
+
- Except for the last tar file, each tar file contains about 32768 audio-text pairs (OGG and TXT files), hence about 65536 files in total (the number may be smaller than 32768 since I removed some files after the initial upload).
|
| 122 |
|
| 123 |
- File names are randomly generated SHA-256 hashes, so the order of the files has no mean (e.g., the files coming from the same Galgame are not necessarily adjacent).
|
| 124 |
|