Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -36,6 +36,14 @@ We use [mesolitica/nanot5-base-malaysian-translation-v2.1](https://huggingface.c
|
|
| 36 |
10. [Malay Conversational Speech Corpus](https://huggingface.co/datasets/malaysia-ai/malay-conversational-speech-corpus-whisper-format)
|
| 37 |
11. [Iban](https://huggingface.co/datasets/malaysia-ai/iban-whisper-format)
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
## how to prepare the dataset
|
| 40 |
|
| 41 |
```bash
|
|
|
|
| 36 |
10. [Malay Conversational Speech Corpus](https://huggingface.co/datasets/malaysia-ai/malay-conversational-speech-corpus-whisper-format)
|
| 37 |
11. [Iban](https://huggingface.co/datasets/malaysia-ai/iban-whisper-format)
|
| 38 |
|
| 39 |
+
### Pseudo-streaming
|
| 40 |
+
|
| 41 |
+
Because Whisper been trained on 30 seconds of audio chunk, so we,
|
| 42 |
+
|
| 43 |
+
1. convert [Malaysian context v2](https://huggingface.co/datasets/mesolitica/pseudolabel-malaysian-youtube-whisper-large-v3-timestamp) into pseudo-streaming chunks, [malaysian_context_v2_chunk](https://huggingface.co/datasets/mesolitica/Malaysian-STT-Whisper/viewer/default/malaysian_context_v2_chunk)
|
| 44 |
+
|
| 45 |
+
**Extremenly large dataset**.
|
| 46 |
+
|
| 47 |
## how to prepare the dataset
|
| 48 |
|
| 49 |
```bash
|