|
--- |
|
license: mit |
|
language: |
|
- en |
|
- pai |
|
metrics: |
|
- bleu |
|
- chrf |
|
datasets: |
|
- hunterschep/paiwan-english |
|
--- |
|
|
|
# Model Performance |
|
|
|
The performance of the translation model has been evaluated using the following metrics: |
|
|
|
- **BLEU Score**: 5.94 |
|
- **chrF++ Score**: 34.99 |
|
|
|
This model was trained on a ~30,000 line parralel-corpus of Paiwan-English translations. |
|
Most of these translations were assembled using Mandarin Chinese as a pivot language. |
|
The model itself was evaluated on the 'test' portion of the dataset. |
|
Total training length was 40,000 epochs. |