Cong123779/AI2Text-Bilingual-ASR-Dataset
收藏Hugging Face2026-02-23 更新2026-03-29 收录
下载链接：
https://hf-mirror.com/datasets/Cong123779/AI2Text-Bilingual-ASR-Dataset
下载链接
链接失效反馈官方服务：
资源简介：
---
language:
- vi
- en
license: cc-by-4.0
task_categories:
- automatic-speech-recognition
pretty_name: AI2Text Bilingual ASR Dataset (Vietnamese + English)
size_categories:
- 100K<n<1M
---
# AI2Text – Bilingual ASR Dataset
A large-scale bilingual (Vietnamese + English) speech dataset used to train
the `Cong123779/AI2Text-Bilingual-ASR` model.
## Dataset Summary
| Split | Samples | Notes |
|-------|---------|-------|
| train | ~194,167 | 77% Vietnamese, 23% English |
| val | ~30,123 | held-out validation |
## Data Fields
Each `manifest.csv` has the following columns:
| Column | Description |
|--------|-------------|
| `id` | Unique sample identifier |
| `transcript` | Ground-truth text (prefixed with `<\|vi\|>` or `<\|en\|>`) |
| `audio_path` | Relative path to the `.wav` file |
| `duration` | Duration in seconds |
| `words_json` | JSON array of word-level timestamps |
## Audio Format
- Sample rate: **16,000 Hz**
- Channels: **Mono**
- Format: **WAV**
## Language Distribution
- **Vietnamese** (`<|vi|>` token): ~77%
- **English** (`<|en|>` token): ~23%
## Usage
```python
import pandas as pd
from datasets import load_dataset
# Load only the manifest CSVs (fast, no audio)
ds = load_dataset("Cong123779/AI2Text-Bilingual-ASR-Dataset",
data_files={"train": "train/manifest.csv",
"val": "val/manifest.csv"})
print(ds)
```
## License
Creative Commons Attribution 4.0 (CC-BY 4.0)
提供机构：
Cong123779
