Skip to content

Latest commit

 

History

History

trk-eng

README.md

opus-2020-06-28.zip

  • dataset: opus
  • model: transformer
  • source language(s): aze_Latn bak chv crh crh_Latn kaz_Cyrl kaz_Latn kir_Cyrl kjh kum ota_Arab ota_Latn sah tat tat_Arab tat_Latn tuk tuk_Latn tur tyv uig_Arab uig_Cyrl uzb_Cyrl uzb_Latn
  • target language(s): eng
  • model: transformer
  • pre-processing: normalization + SentencePiece (spm32k,spm32k)
  • download: opus-2020-06-28.zip
  • test set translations: opus-2020-06-28.test.txt
  • test set scores: opus-2020-06-28.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.aze-eng.aze.eng 28.7 0.472
Tatoeba-test.bak-eng.bak.eng 6.2 0.215
Tatoeba-test.chv-eng.chv.eng 2.0 0.177
Tatoeba-test.crh-eng.crh.eng 14.6 0.352
Tatoeba-test.kaz-eng.kaz.eng 12.2 0.307
Tatoeba-test.kir-eng.kir.eng 16.7 0.395
Tatoeba-test.kjh-eng.kjh.eng 2.4 0.148
Tatoeba-test.kum-eng.kum.eng 18.4 0.325
Tatoeba-test.multi.eng 22.9 0.414
Tatoeba-test.ota-eng.ota.eng 2.6 0.151
Tatoeba-test.sah-eng.sah.eng 0.9 0.126
Tatoeba-test.tat-eng.tat.eng 6.5 0.247
Tatoeba-test.tuk-eng.tuk.eng 6.6 0.264
Tatoeba-test.tur-eng.tur.eng 41.1 0.581
Tatoeba-test.tyv-eng.tyv.eng 3.0 0.173
Tatoeba-test.uig-eng.uig.eng 2.2 0.177
Tatoeba-test.uzb-eng.uzb.eng 9.6 0.248

opus2m-2020-08-01.zip

  • dataset: opus2m
  • model: transformer
  • source language(s): aze_Latn bak chv crh crh_Latn kaz_Cyrl kaz_Latn kir_Cyrl kjh kum ota_Arab ota_Latn sah tat tat_Arab tat_Latn tuk tuk_Latn tur tyv uig_Arab uig_Cyrl uzb_Cyrl uzb_Latn
  • target language(s): eng
  • model: transformer
  • pre-processing: normalization + SentencePiece (spm32k,spm32k)
  • download: opus2m-2020-08-01.zip
  • test set translations: opus2m-2020-08-01.test.txt
  • test set scores: opus2m-2020-08-01.eval.txt

Benchmarks

testset BLEU chr-F
newsdev2016-entr-tureng.tur.eng 5.0 0.242
newstest2016-entr-tureng.tur.eng 3.7 0.231
newstest2017-entr-tureng.tur.eng 3.7 0.229
newstest2018-entr-tureng.tur.eng 4.1 0.230
Tatoeba-test.aze-eng.aze.eng 15.1 0.330
Tatoeba-test.bak-eng.bak.eng 3.3 0.185
Tatoeba-test.chv-eng.chv.eng 1.3 0.161
Tatoeba-test.crh-eng.crh.eng 10.8 0.325
Tatoeba-test.kaz-eng.kaz.eng 9.6 0.264
Tatoeba-test.kir-eng.kir.eng 15.3 0.328
Tatoeba-test.kjh-eng.kjh.eng 1.8 0.121
Tatoeba-test.kum-eng.kum.eng 16.1 0.277
Tatoeba-test.multi.eng 12.0 0.304
Tatoeba-test.ota-eng.ota.eng 2.0 0.149
Tatoeba-test.sah-eng.sah.eng 0.7 0.140
Tatoeba-test.tat-eng.tat.eng 4.0 0.215
Tatoeba-test.tuk-eng.tuk.eng 5.5 0.243
Tatoeba-test.tur-eng.tur.eng 26.8 0.443
Tatoeba-test.tyv-eng.tyv.eng 1.3 0.111
Tatoeba-test.uig-eng.uig.eng 0.2 0.111
Tatoeba-test.uzb-eng.uzb.eng 4.6 0.195

opus1m-2021-02-23.zip

Benchmarks

testset BLEU chr-F #sent #words BP
newsdev2016-entr.tur-eng 9.0 0.316 1001 22011 0.890
newstest2016-entr.tur-eng 7.5 0.300 3000 66175 0.915
newstest2017-entr.tur-eng 7.6 0.300 3007 67703 0.910
newstest2018-entr.tur-eng 8.2 0.304 3000 68725 0.914
Tatoeba-test.aze-eng 6.3 0.234 2659 16165 1.000
Tatoeba-test.bak-eng 4.5 0.217 39 212 1.000
Tatoeba-test.chv-eng 1.2 0.173 333 2240 1.000
Tatoeba-test.crh-eng 16.2 0.295 22 116 1.000
Tatoeba-test.kaz-eng 10.6 0.291 397 2655 1.000
Tatoeba-test.kir-eng 15.8 0.349 118 704 1.000
Tatoeba-test.kjh-eng 0.8 0.111 17 78 1.000
Tatoeba-test.kum-eng 10.1 0.297 8 49 1.000
Tatoeba-test.multi-eng 21.2 0.404 10000 75506 1.000
Tatoeba-test.nog-eng 6.9 0.233 83 459 1.000
Tatoeba-test.ota-eng 1.9 0.167 678 4344 1.000
Tatoeba-test.sah-eng 0.5 0.130 39 225 1.000
Tatoeba-test.tat-eng 5.9 0.241 1451 11261 1.000
Tatoeba-test.tuk-eng 5.8 0.255 2500 19066 1.000
Tatoeba-test.tur-eng 40.7 0.581 10000 77701 0.960
Tatoeba-test.tyv-eng 1.9 0.100 5 25 1.000
Tatoeba-test.uig-eng 1.6 0.171 3024 23090 1.000
Tatoeba-test.uzb-eng 7.1 0.226 457 2699 1.000