---
license: unknown
task_categories:
- automatic-speech-recognition
language:
- en
- ms
- ta
- zh
pretty_name: TIL 26 ASR Stratified Split
---

# TIL 26 ASR Stratified Split

Source data was staged from `/home/jupyter/advanced/asr` without modifying the source folder.
The split is stratified by `language` with seed `3407` and validation ratio `0.2`.

## Files

- `train/train.jsonl` with audio under `train/audio/`
- `val/val.jsonl` with audio under `val/audio/`

## Counts

- Train: 3595
- Val: 899

## Class Counts

| Split | Class | Count |
| --- | --- | ---: |
| train | chinese | 900 |
| train | english | 897 |
| train | malay | 900 |
| train | tamil | 898 |
| val | chinese | 225 |
| val | english | 224 |
| val | malay | 225 |
| val | tamil | 225 |