| #### Languages: | |
| - Source language: English | |
| - Source language: isiZulu | |
| #### Model Details: | |
| - model: transformer | |
| - Architecture: MarianMT | |
| - pre-processing: normalization + SentencePiece | |
| #### Pre-trained Model: | |
| - https://huggingface.co/Helsinki-NLP/opus-mt-en-xh | |
| #### Corpus: | |
| - Umsuka English-isiZulu Parallel Corpus (https://zenodo.org/record/5035171#.Yh5NIOhBy3A) | |
| #### Benchmark: | |
| | Benchmark | Train | Test | | |
| |-----------|-------|-------| | |
| | Umsuka | 17.61 | 13.73 | | |
| #### GitHub: | |
| - https://github.com/umair-nasir14/Geographical-Distance-Is-The-New-Hyperparameter | |
| #### Citation: | |
| ``` | |
| @article{umair2022geographical, | |
| title={Geographical Distance Is The New Hyperparameter: A Case Study Of Finding The Optimal Pre-trained Language For English-isiZulu Machine Translation}, | |
| author={Umair Nasir, Muhammad and Amos Mchechesi, Innocent}, | |
| journal={arXiv e-prints}, | |
| pages={arXiv--2205}, | |
| year={2022} | |
| } | |
| ``` | |