Skip to content

Latest commit

 

History

History
89 lines (69 loc) · 3.96 KB

README.md

File metadata and controls

89 lines (69 loc) · 3.96 KB

opus-2021-02-11.zip

  • dataset: opus
  • model: transformer
  • source language(s): chm est fin fkv hun izh kom krl liv mdf myv olo sma sme udm vep
  • target language(s): bel orv rue rus ukr
  • 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-2021-02-11.zip
  • test set translations: opus-2021-02-11.test.txt
  • test set scores: opus-2021-02-11.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.chm-rus.chm.rus 1.6 0.148
Tatoeba-test.est-bel.est.bel 3.2 0.437
Tatoeba-test.est-rus.est.rus 49.5 0.687
Tatoeba-test.est-ukr.est.ukr 34.5 0.503
Tatoeba-test.fin-rus.fin.rus 42.0 0.630
Tatoeba-test.fin-ukr.fin.ukr 37.4 0.572
Tatoeba-test.hun-bel.hun.bel 21.6 0.433
Tatoeba-test.hun-rus.hun.rus 38.4 0.592
Tatoeba-test.hun-ukr.hun.ukr 37.5 0.588
Tatoeba-test.izh-rus.izh.rus 23.0 0.568
Tatoeba-test.multi.multi 28.7 0.456
Tatoeba-test.udm-rus.udm.rus 1.2 0.182

opus-2021-02-17.zip

  • dataset: opus
  • model: transformer
  • source language(s): est fin hun izh mhr udm
  • target language(s): bel rus ukr
  • 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: >>rus<< >>ukr<< >>bel<< >>bel_Latn<<
  • download: opus-2021-02-17.zip
  • test set translations: opus-2021-02-17.test.txt
  • test set scores: opus-2021-02-17.eval.txt

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test.multi-multi 28.7 0.456 10000 64321 0.979

opus-2021-02-18.zip

  • dataset: opus
  • model: transformer
  • source language(s): est fin hun izh mhr udm
  • target language(s): bel rus ukr
  • 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: >>rus<< >>ukr<< >>bel<< >>bel_Latn<<
  • download: opus-2021-02-18.zip
  • test set translations: opus-2021-02-18.test.txt
  • test set scores: opus-2021-02-18.eval.txt

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test.chm-rus 1.7 0.148 2750 20215 0.961
Tatoeba-test.est-bel 3.2 0.437 2 32 0.867
Tatoeba-test.est-rus 48.2 0.683 598 3563 0.985
Tatoeba-test.est-ukr 34.5 0.503 16 88 0.759
Tatoeba-test.fin-rus 41.0 0.626 3643 21434 1.000
Tatoeba-test.fin-ukr 38.1 0.578 32 209 0.961
Tatoeba-test.hun-bel 21.6 0.433 3 15 1.000
Tatoeba-test.hun-rus 37.6 0.589 2500 16105 0.977
Tatoeba-test.hun-ukr 37.0 0.587 464 2546 0.967
Tatoeba-test.izh-rus 23.0 0.568 8 32 1.000
Tatoeba-test.multi-multi 28.7 0.456 10000 64321 0.979
Tatoeba-test.udm-rus 1.2 0.180 54 317 1.000