- dataset: opus4m+btTCv20210807
- model: transformer
- source language(s): ara arq arz heb
- target language(s): fin
- raw source language(s): ara arq arz heb
- raw target language(s): fin
- model: transformer
- pre-processing: normalization + SentencePiece (spm32k,spm32k)
- download: opus4m+btTCv20210807-2022-01-19.zip
- test set translations: opus4m+btTCv20210807-2022-01-19.test.txt
- test set scores: opus4m+btTCv20210807-2022-01-19.eval.txt
testset | BLEU | chr-F | #sent | #words | BP |
---|---|---|---|---|---|
Tatoeba-test-v2021-08-07.ara-fin | 63.0 | 0.796 | 7 | 34 | 1.000 |
Tatoeba-test-v2021-08-07.heb-fin | 39.4 | 0.630 | 212 | 1302 | 0.944 |
Tatoeba-test-v2021-08-07.multi-fin | 40.2 | 0.635 | 219 | 1336 | 0.948 |