Skip to content

Latest commit

 

History

History
92 lines (72 loc) · 4.14 KB

README.md

File metadata and controls

92 lines (72 loc) · 4.14 KB

opus-2020-07-27.zip

  • dataset: opus
  • model: transformer
  • source language(s): bul mkd srp_Cyrl
  • target language(s): bul mkd srp_Cyrl
  • model: transformer
  • pre-processing: normalization + SentencePiece (spm32k,spm32k)
  • a sentence initial language token is required in the form of >>id<< (id = valid target language ID)
  • download: opus-2020-07-27.zip
  • test set translations: opus-2020-07-27.test.txt
  • test set scores: opus-2020-07-27.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.bul-hbs.bul.hbs 19.3 0.514
Tatoeba-test.bul-mkd.bul.mkd 31.9 0.669
Tatoeba-test.hbs-bul.hbs.bul 18.0 0.636
Tatoeba-test.hbs-mkd.hbs.mkd 19.4 0.322
Tatoeba-test.mkd-bul.mkd.bul 44.6 0.679
Tatoeba-test.mkd-hbs.mkd.hbs 5.5 0.152
Tatoeba-test.multi.multi 26.5 0.563

opus-2020-10-04.zip

  • dataset: opus
  • model: transformer
  • source language(s): bos_Cyrl bos_Latn bul cnr cnr_Latn eng hbs hbs_Cyrl hrv mkd slv srp_Cyrl srp_Latn
  • target language(s): bos_Cyrl bos_Latn bul cnr cnr_Latn eng hbs hbs_Cyrl hrv mkd slv srp_Cyrl srp_Latn
  • model: transformer
  • pre-processing: normalization + SentencePiece (spm32k,spm32k)
  • a sentence initial language token is required in the form of >>id<< (id = valid target language ID)
  • download: opus-2020-10-04.zip
  • test set translations: opus-2020-10-04.test.txt
  • test set scores: opus-2020-10-04.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.bul-eng.bul.eng 53.2 0.682
Tatoeba-test.bul-hbs.bul.hbs 53.7 0.871
Tatoeba-test.bul-mkd.bul.mkd 37.5 0.688
Tatoeba-test.eng-bul.eng.bul 45.6 0.638
Tatoeba-test.eng-hbs.eng.hbs 38.7 0.600
Tatoeba-test.eng-mkd.eng.mkd 43.5 0.627
Tatoeba-test.eng-slv.eng.slv 17.1 0.343
Tatoeba-test.hbs-bul.hbs.bul 100.0 1.000
Tatoeba-test.hbs-eng.hbs.eng 52.9 0.680
Tatoeba-test.hbs-mkd.hbs.mkd 38.0 0.718
Tatoeba-test.mkd-bul.mkd.bul 54.2 0.788
Tatoeba-test.mkd-eng.mkd.eng 52.7 0.667
Tatoeba-test.mkd-hbs.mkd.hbs 53.7 0.871
Tatoeba-test.multi.multi 44.2 0.596
Tatoeba-test.slv-eng.slv.eng 24.8 0.403

opus-2021-02-23.zip

  • dataset: opus
  • model: transformer
  • source language(s): bos bul hbs hrv mkd srp
  • target language(s): bos bul hbs hrv mkd srp
  • model: transformer
  • pre-processing: normalization + SentencePiece (spm32k,spm32k)
  • a sentence initial language token is required in the form of >>id<< (id = valid target language ID)
  • valid language labels: >>eng<< >>mkd<< >>bul<< >>srp_Cyrl<< >>srp_Latn<< >>bos_Latn<< >>hrv<< >>hbs<< >>cnr_Latn<< >>slv<< >>hbs_Cyrl<< >>bos_Cyrl<< >>cnr<<
  • download: opus-2021-02-23.zip
  • test set translations: opus-2021-02-23.test.txt
  • test set scores: opus-2021-02-23.eval.txt

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test.bul-hbs 53.7 0.871 1 6 1.000
Tatoeba-test.bul-mkd 37.5 0.688 3 17 1.000
Tatoeba-test.hbs-bul 100.0 1.000 1 6 1.000
Tatoeba-test.hbs-mkd 38.0 0.718 1 6 1.000
Tatoeba-test.mkd-bul 54.2 0.788 3 17 1.000
Tatoeba-test.mkd-hbs 53.7 0.871 1 6 1.000
Tatoeba-test.multi-multi 48.3 0.687 14 82 1.000