Harnessing transfer learning for low-resource language translation
# English-Zulu-MarianMT-Model
This repo contains nine files containing the framework necessary to fine-tune and evaluate pre-trained MarianMT models on the Umsuka English-isiZulu Parallel Corpus.
Selected Models:
- En-xh
- En-sw
- Multilingual Romance
Selected tokenization schemes:
- Byte Pair Encoding
- WordPiece
- SentencePiece
## Report
Full report here:
Effective Transfer Learning Between Morphologically Similar Languages:
A Case Study on English-Zulu Translation