Skip to content

Latest commit

 

History

History

trk-fiu

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
README.md
README.md
 
 
opus-2021-02-19.yml
opus-2021-02-19.yml
 
 
opus.yml
opus.yml
 
 

README.md

opus-2021-02-10.zip

  • dataset: opus
  • model: transformer
  • source language(s): alt aze bak chg chv crh gag kaa kaz kir kjh krc kum nog ota otk sah tat tuk tur tyv uig uzb
  • target language(s): chm est fin fkv hun izh kom krl liv mdf myv olo sma sme udm vep
  • 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-2021-02-10.zip
  • test set translations: opus-2021-02-10.test.txt
  • test set scores: opus-2021-02-10.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.bak-est.bak.est 4.0 0.161
Tatoeba-test.kaz-est.kaz.est 32.5 0.439
Tatoeba-test.multi.multi 26.4 0.513
Tatoeba-test.ota-hun.ota.hun 26.9 0.231
Tatoeba-test.tat-chm.tat.chm 2.7 0.005
Tatoeba-test.tat-est.tat.est 4.7 0.271
Tatoeba-test.tat-fin.tat.fin 3.6 0.225
Tatoeba-test.tat-hun.tat.hun 13.5 0.333
Tatoeba-test.tur-est.tur.est 42.5 0.563
Tatoeba-test.tur-fin.tur.fin 25.9 0.512
Tatoeba-test.tur-hun.tur.hun 27.7 0.524
Tatoeba-test.uzb-est.uzb.est 35.0 0.435

opus-2021-02-19.zip

  • dataset: opus
  • model: transformer
  • source language(s): bak kaz ota tat tur uzb
  • target language(s): est fin hun mhr
  • 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: >>est<< >>fin<< >>hun<<
  • download: opus-2021-02-19.zip
  • test set translations: opus-2021-02-19.test.txt
  • test set scores: opus-2021-02-19.eval.txt

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test.bak-est 4.0 0.161 1 6 1.000
Tatoeba-test.kaz-est 32.5 0.439 1 6 1.000
Tatoeba-test.multi-multi 26.4 0.513 2798 15728 0.942
Tatoeba-test.ota_Arab-hun 0.0 0.000 1 3 0.607
Tatoeba-test.ota-hun 26.9 0.231 2 6 1.000
Tatoeba-test.ota_Latn-hun 31.9 0.407 1 3 1.000
Tatoeba-test.tat-chm 2.7 0.005 1 12 1.000
Tatoeba-test.tat-est 4.7 0.271 3 17 0.939
Tatoeba-test.tat-fin 3.6 0.225 6 83 0.568
Tatoeba-test.tat-hun 4.0 0.265 1 8 1.000
Tatoeba-test.tat_Latn-est 10.7 0.213 1 5 1.000
Tatoeba-test.tat_Latn-fin 1.0 0.110 1 30 0.178
Tatoeba-test.tur-est 42.5 0.563 21 117 0.873
Tatoeba-test.tur-fin 25.9 0.512 1796 10455 0.918
Tatoeba-test.tur-hun 27.7 0.524 965 4994 0.998
Tatoeba-test.uzb-est 35.0 0.435 1 6 0.819