eng-deu+eng+fin+fra+nld+swe

opus1m-2021-02-22.zip

• dataset: opus1m
• model: transformer-align
• source language(s): eng
• target language(s): deu fin fra nld swe
• model: transformer-align
• 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: >>deu<< >>fra<< >>swe<< >>fin<< >>nld<<
• download: opus1m-2021-02-22.zip
• test set translations: opus1m-2021-02-22.test.txt
• test set scores: opus1m-2021-02-22.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.eng-fin	18.1	0.518	1500	23375	1.000
newsdiscussdev2015-enfr.eng-fra	31.3	0.582	1500	27986	1.000
newsdiscusstest2015-enfr.eng-fra	36.7	0.617	1500	28027	0.996
newssyscomb2009.eng-deu	21.1	0.521	502	11271	0.985
newssyscomb2009.eng-fra	28.2	0.570	502	12334	0.999
news-test2008.eng-deu	21.2	0.512	2051	47427	0.992
news-test2008.eng-fra	25.2	0.539	2051	52685	0.995
newstest2009.eng-deu	20.7	0.514	2525	62816	0.991
newstest2009.eng-fra	27.2	0.561	2525	69278	0.980
newstest2010.eng-deu	22.8	0.527	2489	61511	0.956
newstest2010.eng-fra	30.0	0.579	2489	66043	0.988
newstest2011.eng-deu	20.7	0.509	3003	72981	0.980
newstest2011.eng-fra	31.5	0.593	3003	80626	0.968
newstest2012.eng-deu	21.2	0.509	3003	72886	0.959
newstest2012.eng-fra	29.6	0.577	3003	78011	0.982
newstest2013.eng-deu	24.4	0.534	3000	63737	0.972
newstest2013.eng-fra	30.6	0.570	3000	70037	0.974
newstest2015-ende.eng-deu	28.6	0.572	2169	44260	1.000
newstest2015-enfi.eng-fin	19.5	0.528	1370	19968	1.000
newstest2016-ende.eng-deu	33.3	0.608	2999	62670	0.986
newstest2016-enfi.eng-fin	19.9	0.536	3000	48116	0.976
newstest2017-ende.eng-deu	27.1	0.562	3004	61291	1.000
newstest2017-enfi.eng-fin	22.5	0.561	3002	45718	0.985
newstest2018-ende.eng-deu	39.6	0.650	2998	64276	0.995
newstest2018-enfi.eng-fin	15.3	0.496	3000	45475	1.000
newstest2019-ende.eng-deu	35.5	0.616	1997	48969	0.971
newstest2019-enfi.eng-fin	19.7	0.516	1997	38369	0.946
newstestB2016-enfi.eng-fin	16.0	0.505	3000	45766	1.000
newstestB2017-enfi.eng-fin	18.5	0.524	3002	45506	0.990
Tatoeba-test.eng-deu	40.4	0.614	10000	83347	0.985
Tatoeba-test.eng-fin	33.0	0.583	10000	60517	0.925
Tatoeba-test.eng-fra	45.4	0.636	10000	80769	0.974
Tatoeba-test.eng-multi	46.1	0.643	10000	72342	0.967
Tatoeba-test.eng-nld	52.8	0.697	10000	71436	0.976
Tatoeba-test.eng-swe	56.7	0.710	10000	65581	0.963
tico19-test.eng-fra	37.3	0.614	2100	64655	0.988