Skip to content

Latest commit

 

History

History

gmq-gmq

README.md

opus-2020-07-06.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan fao isl nno nob swe
  • target language(s): dan fao isl nno nob swe
  • 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-2020-07-06.zip
  • test set translations: opus-2020-07-06.test.txt
  • test set scores: opus-2020-07-06.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.dan-fao.dan.fao 8.1 0.122
Tatoeba-test.dan-isl.dan.isl 70.7 0.909
Tatoeba-test.dan-nor.dan.nor 12.1 0.417
Tatoeba-test.dan-swe.dan.swe 67.5 0.800
Tatoeba-test.fao-dan.fao.dan 18.0 0.474
Tatoeba-test.fao-isl.fao.isl 11.3 0.262
Tatoeba-test.fao-nor.fao.nor 7.9 0.299
Tatoeba-test.fao-swe.fao.swe 35.4 0.830
Tatoeba-test.isl-dan.isl.dan 100.0 1.000
Tatoeba-test.isl-fao.isl.fao 14.5 0.206
Tatoeba-test.isl-nor.isl.nor 14.1 0.396
Tatoeba-test.isl-swe.isl.swe 73.5 0.793
Tatoeba-test.multi.multi 67.5 0.799
Tatoeba-test.nor-dan.nor.dan 52.9 0.713
Tatoeba-test.nor-fao.nor.fao 2.0 0.228
Tatoeba-test.nor-isl.nor.isl 20.8 0.453
Tatoeba-test.nor-nor.nor.nor 11.2 0.419
Tatoeba-test.nor-swe.nor.swe 52.7 0.701
Tatoeba-test.swe-dan.swe.dan 67.5 0.799
Tatoeba-test.swe-fao.swe.fao 0.0 0.279
Tatoeba-test.swe-isl.swe.isl 70.7 0.822
Tatoeba-test.swe-nor.swe.nor 23.4 0.519

opus-2020-07-21.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan fao isl nno nob swe
  • target language(s): dan fao isl nno nob swe
  • 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-2020-07-21.zip
  • test set translations: opus-2020-07-21.test.txt
  • test set scores: opus-2020-07-21.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.dan-fao.dan.fao 8.1 0.173
Tatoeba-test.dan-isl.dan.isl 52.5 0.827
Tatoeba-test.dan-nor.dan.nor 62.8 0.769
Tatoeba-test.dan-swe.dan.swe 67.5 0.802
Tatoeba-test.fao-dan.fao.dan 14.5 0.255
Tatoeba-test.fao-isl.fao.isl 26.3 0.359
Tatoeba-test.fao-nor.fao.nor 36.5 0.462
Tatoeba-test.fao-swe.fao.swe 0.0 0.632
Tatoeba-test.isl-dan.isl.dan 67.0 0.739
Tatoeba-test.isl-fao.isl.fao 14.5 0.226
Tatoeba-test.isl-nor.isl.nor 50.2 0.650
Tatoeba-test.isl-swe.isl.swe 100.0 1.000
Tatoeba-test.multi.multi 64.9 0.783
Tatoeba-test.nor-dan.nor.dan 66.0 0.800
Tatoeba-test.nor-fao.nor.fao 9.9 0.345
Tatoeba-test.nor-isl.nor.isl 38.5 0.588
Tatoeba-test.nor-nor.nor.nor 52.4 0.727
Tatoeba-test.nor-swe.nor.swe 67.2 0.796
Tatoeba-test.swe-dan.swe.dan 68.0 0.803
Tatoeba-test.swe-fao.swe.fao 0.0 0.268
Tatoeba-test.swe-isl.swe.isl 32.5 0.623
Tatoeba-test.swe-nor.swe.nor 61.6 0.763

opus-2020-07-27.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan fao isl nno nob swe
  • target language(s): dan fao isl nno nob swe
  • 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.dan-fao.dan.fao 8.1 0.173
Tatoeba-test.dan-isl.dan.isl 52.5 0.827
Tatoeba-test.dan-nor.dan.nor 62.8 0.772
Tatoeba-test.dan-swe.dan.swe 67.6 0.802
Tatoeba-test.fao-dan.fao.dan 11.3 0.306
Tatoeba-test.fao-isl.fao.isl 26.3 0.359
Tatoeba-test.fao-nor.fao.nor 36.8 0.531
Tatoeba-test.fao-swe.fao.swe 0.0 0.632
Tatoeba-test.isl-dan.isl.dan 67.0 0.739
Tatoeba-test.isl-fao.isl.fao 14.5 0.243
Tatoeba-test.isl-nor.isl.nor 51.8 0.674
Tatoeba-test.isl-swe.isl.swe 100.0 1.000
Tatoeba-test.multi.multi 64.7 0.782
Tatoeba-test.nor-dan.nor.dan 65.6 0.797
Tatoeba-test.nor-fao.nor.fao 9.4 0.362
Tatoeba-test.nor-isl.nor.isl 38.8 0.587
Tatoeba-test.nor-nor.nor.nor 51.9 0.721
Tatoeba-test.nor-swe.nor.swe 66.5 0.789
Tatoeba-test.swe-dan.swe.dan 67.6 0.802
Tatoeba-test.swe-fao.swe.fao 0.0 0.268
Tatoeba-test.swe-isl.swe.isl 65.8 0.914
Tatoeba-test.swe-nor.swe.nor 60.6 0.755

opus-2020-09-26.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan eng fao isl nno nob nob_Hebr non_Latn swe
  • target language(s): dan eng fao isl nno nob nob_Hebr non_Latn swe
  • 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-2020-09-26.zip
  • test set translations: opus-2020-09-26.test.txt
  • test set scores: opus-2020-09-26.eval.txt

Benchmarks

testset BLEU chr-F
Tatoeba-test.dan-eng.dan.eng 59.1 0.737
Tatoeba-test.dan-fao.dan.fao 6.6 0.162
Tatoeba-test.dan-isl.dan.isl 65.8 0.914
Tatoeba-test.dan-nor.dan.nor 74.3 0.849
Tatoeba-test.dan-swe.dan.swe 69.5 0.818
Tatoeba-test.eng-dan.eng.dan 57.0 0.717
Tatoeba-test.eng-fao.eng.fao 8.3 0.305
Tatoeba-test.eng-isl.eng.isl 24.2 0.505
Tatoeba-test.eng-non.eng.non 0.3 0.160
Tatoeba-test.eng-nor.eng.nor 51.7 0.681
Tatoeba-test.eng-swe.eng.swe 57.3 0.710
Tatoeba-test.fao-dan.fao.dan 23.0 0.536
Tatoeba-test.fao-eng.fao.eng 26.0 0.454
Tatoeba-test.fao-isl.fao.isl 19.0 0.303
Tatoeba-test.fao-nor.fao.nor 47.8 0.631
Tatoeba-test.fao-swe.fao.swe 0.0 1.000
Tatoeba-test.isl-dan.isl.dan 77.7 0.901
Tatoeba-test.isl-eng.isl.eng 49.6 0.659
Tatoeba-test.isl-fao.isl.fao 14.5 0.183
Tatoeba-test.isl-nor.isl.nor 60.5 0.761
Tatoeba-test.isl-swe.isl.swe 60.0 0.800
Tatoeba-test.multi.multi 57.0 0.716
Tatoeba-test.non-eng.non.eng 26.7 0.501
Tatoeba-test.nor-dan.nor.dan 71.5 0.834
Tatoeba-test.nor-eng.nor.eng 54.2 0.694
Tatoeba-test.nor-fao.nor.fao 13.1 0.377
Tatoeba-test.nor-isl.nor.isl 29.7 0.545
Tatoeba-test.nor-nor.nor.nor 61.9 0.786
Tatoeba-test.nor-swe.nor.swe 72.9 0.838
Tatoeba-test.swe-dan.swe.dan 68.7 0.809
Tatoeba-test.swe-eng.swe.eng 60.9 0.737
Tatoeba-test.swe-fao.swe.fao 12.7 0.290
Tatoeba-test.swe-isl.swe.isl 65.8 0.914
Tatoeba-test.swe-nor.swe.nor 72.0 0.832

opus-2020-10-04.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan eng fao isl nno nob nob_Hebr non_Latn swe
  • target language(s): dan eng fao isl nno nob nob_Hebr non_Latn swe
  • 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.dan-eng.dan.eng 59.2 0.738
Tatoeba-test.dan-fao.dan.fao 6.6 0.171
Tatoeba-test.dan-isl.dan.isl 46.7 0.758
Tatoeba-test.dan-nor.dan.nor 74.1 0.849
Tatoeba-test.dan-swe.dan.swe 69.3 0.818
Tatoeba-test.eng-dan.eng.dan 57.1 0.719
Tatoeba-test.eng-fao.eng.fao 8.9 0.315
Tatoeba-test.eng-isl.eng.isl 24.3 0.506
Tatoeba-test.eng-non.eng.non 0.4 0.182
Tatoeba-test.eng-nor.eng.nor 51.9 0.682
Tatoeba-test.eng-swe.eng.swe 57.3 0.712
Tatoeba-test.fao-dan.fao.dan 23.0 0.536
Tatoeba-test.fao-eng.fao.eng 25.4 0.453
Tatoeba-test.fao-isl.fao.isl 15.2 0.304
Tatoeba-test.fao-nor.fao.nor 43.9 0.602
Tatoeba-test.fao-swe.fao.swe 0.0 1.000
Tatoeba-test.isl-dan.isl.dan 77.7 0.901
Tatoeba-test.isl-eng.isl.eng 49.7 0.660
Tatoeba-test.isl-fao.isl.fao 14.5 0.207
Tatoeba-test.isl-nor.isl.nor 61.9 0.772
Tatoeba-test.isl-swe.isl.swe 60.0 0.800
Tatoeba-test.multi.multi 57.1 0.717
Tatoeba-test.non-eng.non.eng 30.0 0.523
Tatoeba-test.nor-dan.nor.dan 71.1 0.831
Tatoeba-test.nor-eng.nor.eng 54.2 0.695
Tatoeba-test.nor-fao.nor.fao 9.9 0.343
Tatoeba-test.nor-isl.nor.isl 29.3 0.543
Tatoeba-test.nor-nor.nor.nor 61.0 0.780
Tatoeba-test.nor-swe.nor.swe 72.5 0.839
Tatoeba-test.swe-dan.swe.dan 68.7 0.808
Tatoeba-test.swe-eng.swe.eng 61.1 0.738
Tatoeba-test.swe-fao.swe.fao 19.0 0.164
Tatoeba-test.swe-isl.swe.isl 46.7 0.784
Tatoeba-test.swe-nor.swe.nor 72.9 0.835

opus-2021-02-16.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan eng fao isl nno nob non swe
  • target language(s): dan eng fao isl nno nob non swe
  • 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<< >>isl<< >>swe<< >>dan<< >>nob<< >>nno<< >>fao<<
  • download: opus-2021-02-16.zip
  • test set translations: opus-2021-02-16.test.txt
  • test set scores: opus-2021-02-16.eval.txt

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test.multi-multi 56.3 0.709 10000 72790 0.974

opus-2021-02-18.zip

  • dataset: opus
  • model: transformer
  • source language(s): dan fao isl nno nob swe
  • target language(s): dan fao isl nno nob swe
  • 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<< >>isl<< >>swe<< >>dan<< >>nob<< >>nno<< >>fao<<
  • 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.dan-fao 6.6 0.171 1 6 1.000
Tatoeba-test.dan-isl 46.7 0.758 1 10 1.000
Tatoeba-test.dan-nno 52.8 0.718 12 71 1.000
Tatoeba-test.dan-nob 74.1 0.849 1299 9617 0.996
Tatoeba-test.dan-nor 74.0 0.848 1311 9688 0.996
Tatoeba-test.dan-swe 69.1 0.817 1550 10078 0.988
Tatoeba-test.fao-dan 23.0 0.536 1 6 1.000
Tatoeba-test.fao-isl 15.2 0.304 2 7 1.000
Tatoeba-test.fao-nor 43.9 0.602 21 127 1.000
Tatoeba-test.fao-swe 0.0 1.000 1 3 1.000
Tatoeba-test.isl-dan 77.7 0.901 1 12 0.913
Tatoeba-test.isl-fao 14.5 0.207 2 7 1.000
Tatoeba-test.isl-nor 61.6 0.770 126 921 0.971
Tatoeba-test.isl-swe 60.0 0.800 1 12 1.000
Tatoeba-test.multi-multi 56.3 0.709 10000 72790 0.974
Tatoeba-test.nno-dan 84.1 0.858 12 71 0.971
Tatoeba-test.nno-nob 75.8 0.858 474 3167 0.995
Tatoeba-test.nno-swe 39.8 0.698 2 11 0.905
Tatoeba-test.nob-dan 71.1 0.831 1299 9792 0.998
Tatoeba-test.nob-nno 45.7 0.699 474 3184 1.000
Tatoeba-test.nob-swe 72.1 0.837 560 3660 0.991
Tatoeba-test.nor-dan 71.2 0.831 1311 9863 0.998
Tatoeba-test.nor-fao 9.9 0.343 21 126 0.891
Tatoeba-test.nor-isl 29.3 0.543 126 882 0.975
Tatoeba-test.nor-nor 61.0 0.779 948 6351 1.000
Tatoeba-test.nor-swe 72.1 0.837 562 3671 0.991
Tatoeba-test.swe-dan 68.5 0.808 1550 10260 0.990
Tatoeba-test.swe-fao 19.0 0.164 1 3 1.000
Tatoeba-test.swe-isl 46.7 0.784 1 10 1.000
Tatoeba-test.swe-nno 20.4 0.531 2 10 1.000
Tatoeba-test.swe-nob 72.8 0.834 560 3671 0.991
Tatoeba-test.swe-nor 72.7 0.833 562 3681 0.992

opus4m+btTCv20210807-2021-12-08.zip

  • dataset: opus4m+btTCv20210807
  • model: transformer-big
  • source language(s): dan eng fao isl nno nob nob_Zinh non_Latn swe
  • target language(s): dan eng fao isl nno nob nob_Zinh non_Latn swe
  • raw source language(s): dan eng fao isl nno nob non swe
  • raw target language(s): dan eng fao isl nno nob non swe
  • model: transformer-big
  • 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:
  • download: opus4m+btTCv20210807-2021-12-08.zip
  • test set translations: opus4m+btTCv20210807-2021-12-08.test.txt
  • test set scores: opus4m+btTCv20210807-2021-12-08.eval.txt

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test-v2021-08-07.dan-dan 58.4 0.7420 212 1666 0.989
Tatoeba-test-v2021-08-07.dan-fao 8.1 0.1217 1 6 1.000
Tatoeba-test-v2021-08-07.dan-isl 42.7 0.6289 1 10 1.000
Tatoeba-test-v2021-08-07.dan-nor 76.6 0.8667 1311 9688 1.000
Tatoeba-test-v2021-08-07.dan-swe 69.9 0.8200 1549 10056 0.986
Tatoeba-test-v2021-08-07.fao-dan 14.5 0.2806 1 6 1.000
Tatoeba-test-v2021-08-07.fao-fao 13.9 0.2695 1 8 0.549
Tatoeba-test-v2021-08-07.fao-isl 15.6 0.2929 2 7 1.000
Tatoeba-test-v2021-08-07.fao-nor 53.3 0.6856 21 127 0.984
Tatoeba-test-v2021-08-07.fao-swe 0.0 0.1428 1 3 0.607
Tatoeba-test-v2021-08-07.gmq-gmq 57.2 0.7159 10000 71144 0.984
Tatoeba-test-v2021-08-07.isl-dan 64.3 0.8009 1 12 0.819
Tatoeba-test-v2021-08-07.isl-fao 9.0 0.2780 2 7 1.000
Tatoeba-test-v2021-08-07.isl-nor 60.7 0.7422 126 921 0.957
Tatoeba-test-v2021-08-07.isl-swe 64.3 0.7791 1 12 0.819
Tatoeba-test-v2021-08-07.multi-multi 57.2 0.7159 10000 71144 0.984
Tatoeba-test-v2021-08-07.nor-dan 73.2 0.8487 1311 9863 1.000
Tatoeba-test-v2021-08-07.nor-fao 13.9 0.3376 21 126 0.934
Tatoeba-test-v2021-08-07.nor-isl 36.9 0.5731 126 882 0.984
Tatoeba-test-v2021-08-07.nor-nor 68.1 0.8186 982 6589 0.993
Tatoeba-test-v2021-08-07.nor-swe 71.4 0.8296 566 3735 0.984
Tatoeba-test-v2021-08-07.swe-dan 70.2 0.8216 1549 10238 0.997
Tatoeba-test-v2021-08-07.swe-fao 0.0 0.2847 1 3 1.000
Tatoeba-test-v2021-08-07.swe-isl 70.7 0.8225 1 10 1.000
Tatoeba-test-v2021-08-07.swe-nor 73.1 0.8383 566 3743 0.990
Tatoeba-test-v2021-08-07.swe-swe 46.1 0.6910 1022 6846 0.971

opusTCv20210807_transformer-big_2022-07-29.zip

Benchmarks

testset BLEU chr-F #sent #words BP
Tatoeba-test-v2021-08-07.dan-dan 57.7 0.72908 212 1666 0.999
Tatoeba-test-v2021-08-07.dan-fao 6.6 0.12173 1 6 1.000
Tatoeba-test-v2021-08-07.dan-isl 70.7 0.88923 1 10 1.000
Tatoeba-test-v2021-08-07.dan-nno 45.6 0.71851 12 71 0.986
Tatoeba-test-v2021-08-07.dan-nob 78.3 0.87595 1299 9617 1.000
Tatoeba-test-v2021-08-07.dan-nor 78.1 0.87500 1311 9688 1.000
Tatoeba-test-v2021-08-07.dan-swe 72.2 0.83416 1549 10056 0.987
Tatoeba-test-v2021-08-07.fao-dan 18.0 0.16365 1 6 1.000
Tatoeba-test-v2021-08-07.fao-fao 10.6 0.18204 1 8 1.000
Tatoeba-test-v2021-08-07.fao-isl 69.1 0.71879 2 7 1.000
Tatoeba-test-v2021-08-07.fao-nor 52.5 0.64540 21 127 1.000
Tatoeba-test-v2021-08-07.fao-swe 0.0 10.00000 1 3 1.000
Tatoeba-test-v2021-08-07.isl-dan 73.5 0.80948 1 12 1.000
Tatoeba-test-v2021-08-07.isl-fao 15.6 0.15502 2 7 1.000
Tatoeba-test-v2021-08-07.isl-nor 64.9 0.77509 126 921 0.962
Tatoeba-test-v2021-08-07.isl-swe 64.3 0.77909 1 12 0.819
Tatoeba-test-v2021-08-07.multi-multi 73.3 0.84085 7158 49467 0.995
Tatoeba-test-v2021-08-07.nno-dan 92.1 0.91928 12 71 0.986
Tatoeba-test-v2021-08-07.nno-nob 79.1 0.87845 467 3129 0.998
Tatoeba-test-v2021-08-07.nno-swe 22.2 0.49696 3 38 0.859
Tatoeba-test-v2021-08-07.nob-dan 73.8 0.85277 1299 9792 1.000
Tatoeba-test-v2021-08-07.nob-nno 54.6 0.74545 466 3141 0.995
Tatoeba-test-v2021-08-07.nob-nob 42.9 0.65096 49 319 1.000
Tatoeba-test-v2021-08-07.nob-swe 73.6 0.84567 563 3697 0.988
Tatoeba-test-v2021-08-07.nor-dan 74.0 0.85316 1311 9863 1.000
Tatoeba-test-v2021-08-07.nor-fao 16.8 0.39138 21 126 0.865
Tatoeba-test-v2021-08-07.nor-isl 38.7 0.60109 126 882 0.984
Tatoeba-test-v2021-08-07.nor-nor 66.1 0.80454 982 6589 0.999
Tatoeba-test-v2021-08-07.nor-swe 73.0 0.84102 566 3735 0.987
Tatoeba-test-v2021-08-07.swe-dan 72.0 0.83189 1549 10238 0.994
Tatoeba-test-v2021-08-07.swe-fao 0.0 0.26774 1 3 1.000
Tatoeba-test-v2021-08-07.swe-isl 70.7 0.88923 1 10 1.000
Tatoeba-test-v2021-08-07.swe-nno 7.8 0.47837 3 36 1.000
Tatoeba-test-v2021-08-07.swe-nob 75.8 0.85426 563 3707 0.991
Tatoeba-test-v2021-08-07.swe-nor 75.1 0.84881 566 3743 0.991
Tatoeba-test-v2021-08-07.swe-swe 46.8 0.67091 1022 6846 0.954