Cong123779/mtet-vi-en-bidirectional

---
language:
- vi
- en
license: mit
task_categories:
- translation
tags:
- vietnamese
- english
- parallel-corpus
- bidirectional
size_categories:
- 1M<n<10M
---

# MTET Vietnamese–English Bidirectional Dataset (Cleaned)

This is the cleaned, bidirectional version of the MTET (Machine Translation Evaluation for Translation) Vietnamese–English parallel corpus used to train the [vi-en-transformer-25m](https://huggingface.co/Cong123779/vi-en-transformer-25m) model.

## Dataset Description
- **Source:** MTET corpus (public Vietnamese–English parallel data)
- **Processing:** Cleaned with heuristic filters, bidirectional pairs added (both VI→EN and EN→VI)
- **Format:** CSV with columns `source` and `target`
- **Size:** ~2 GB (cleaned)

## Files
| File | Description |
|---|---|
| `mtet_bidirectional_cleaned.csv` | Main training data (bidirectional, cleaned) |

## Usage
```python
import pandas as pd
df = pd.read_csv("mtet_bidirectional_cleaned.csv")
print(df.head())
# source target
# xin chào hello
# hello xin chào
```

## License
The dataset follows the original MTET distribution license (MIT).