- dataset: opus
- model: transformer
- source language(s): swa swc swh
- target language(s): eng
- model: transformer
- pre-processing: normalization + SentencePiece (spm32k,spm32k)
- download: opus-2021-02-23.zip
- test set translations: opus-2021-02-23.test.txt
- test set scores: opus-2021-02-23.eval.txt
| testset | BLEU | chr-F | #sent | #words | BP |
|---|---|---|---|---|---|
| Tatoeba-test.swa-eng | 43.5 | 0.580 | 772 | 4998 | 0.956 |
| tico19-test.swa-eng | 26.3 | 0.507 | 2100 | 56339 | 1.000 |
- dataset: opus+bt
- model: transformer-align
- source language(s): swa swc swh
- target language(s): eng
- model: transformer-align
- pre-processing: normalization + SentencePiece (spm32k,spm32k)
- download: opus+bt-2021-04-30.zip
- test set translations: opus+bt-2021-04-30.test.txt
- test set scores: opus+bt-2021-04-30.eval.txt
| testset | BLEU | chr-F | #sent | #words | BP |
|---|---|---|---|---|---|
| Tatoeba-test.swa-eng | 47.7 | 0.625 | 772 | 4998 | 0.948 |
| tico19-test.swa-eng | 29.2 | 0.540 | 2100 | 56339 | 1.000 |
- dataset: opusTCv20210807+nopar+ft95
- model: transformer-tiny11-align
- source language(s): swa swh
- target language(s): eng
- raw source language(s): swa swh
- raw target language(s): eng
- model: transformer-tiny11-align
- pre-processing: normalization + SentencePiece (spm32k,spm32k)
- download: opusTCv20210807+nopar+ft95-sepvoc_transformer-tiny11-align_2023-03-13.zip
- test set translations: opusTCv20210807+nopar+ft95-sepvoc_transformer-tiny11-align_2023-03-13.test.txt
- test set scores: opusTCv20210807+nopar+ft95-sepvoc_transformer-tiny11-align_2023-03-13.eval.txt
| testset | BLEU | chr-F | #sent | #words | BP |
|---|---|---|---|---|---|
| Tatoeba-test-v2021-08-07.intgemm8.multi-eng | 48.3 | 0.62864 | 387 | 2508 | 0.928 |
| Tatoeba-test-v2021-08-07.intgemm8.shortlist.multi-eng | 48.3 | 0.62874 | 387 | 2508 | 0.928 |
| Tatoeba-test-v2021-08-07.multi-eng | 49.1 | 0.63415 | 387 | 2508 | 0.932 |
| tico19-test.swa-eng | 28.9 | 0.54057 | 2100 | 56339 | 1.000 |