# English to Tshivenda

Author: Laura Martinus

## Data

	- The JW300 English-Tshivenda dataset.

## Model

- Default Masakhane Transformer translation model.
- [Link to google drive folder with models](https://drive.google.com/open?id=1jeSQB7OLz4pRj_0mLuu9GCLownR2EUwp)

## Analysis
TO DO

# Results

Tokenization | BLEU dev | BLEU test
--- | --- | ---
BPE| 41.14 | 49.57

## BPE
BPE Codes set to 40 000
- [Link to google drive folder with bpes](https://drive.google.com/open?id=1jeSQB7OLz4pRj_0mLuu9GCLownR2EUwp)