eng-fra

opus-2021-02-15.zip

• dataset: opus
• model: transformer-align
• source language(s): eng
• target language(s): fra
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus-2021-02-15.zip
• test set translations: opus-2021-02-15.test.txt
• test set scores: opus-2021-02-15.eval.txt

Benchmarks

testset	BLEU	chr-F
newsdiscussdev2015-enfr-engfra.eng.fra	33.6	0.603
newsdiscusstest2015-enfr-engfra.eng.fra	40.1	0.643
newssyscomb2009-engfra.eng.fra	29.4	0.583
news-test2008-engfra.eng.fra	27.4	0.553
newstest2009-engfra.eng.fra	29.4	0.577
newstest2010-engfra.eng.fra	32.8	0.597
newstest2011-engfra.eng.fra	34.2	0.611
newstest2012-engfra.eng.fra	31.8	0.592
newstest2013-engfra.eng.fra	33.2	0.589
Tatoeba-test.eng.fra	50.5	0.669
Tatoeba-test.eng-fra.eng.fra	50.5	0.669
tico19-test-engfra.eng.fra	41.6	0.638

opus-2021-02-17.zip

• dataset: opus
• model: transformer-align
• source language(s): eng
• target language(s): fra
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• 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.eng-fra	50.7	0.670	10000	80769	0.980

opus-2021-02-19.zip

• dataset: opus
• model: transformer-align
• source language(s): eng
• target language(s): fra
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• 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
newsdiscussdev2015-enfr.eng-fra	33.7	0.604	1500	27986	1.000
newsdiscusstest2015-enfr.eng-fra	40.4	0.645	1500	28027	1.000
newssyscomb2009.eng-fra	30.4	0.588	502	12334	1.000
news-test2008.eng-fra	27.5	0.554	2051	52685	1.000
newstest2009.eng-fra	29.7	0.579	2525	69278	0.990
newstest2010.eng-fra	33.1	0.598	2489	66043	0.996
newstest2011.eng-fra	34.4	0.612	3003	80626	0.977
newstest2012.eng-fra	31.8	0.592	3003	78011	0.988
newstest2013.eng-fra	33.4	0.591	3000	70037	0.980
Tatoeba-test.eng-fra	50.6	0.669	10000	80769	0.981
tico19-test.eng-fra	41.5	0.636	2100	64655	0.999

opus-2021-02-22.zip

• dataset: opus
• model: transformer-align
• source language(s): eng
• target language(s): fra
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus-2021-02-22.zip
• test set translations: opus-2021-02-22.test.txt
• test set scores: opus-2021-02-22.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdiscussdev2015-enfr.eng-fra	33.9	0.606	1500	27986	1.000
newsdiscusstest2015-enfr.eng-fra	40.8	0.649	1500	28027	1.000
newssyscomb2009.eng-fra	30.0	0.587	502	12334	1.000
news-test2008.eng-fra	27.5	0.556	2051	52685	1.000
newstest2009.eng-fra	29.9	0.580	2525	69278	0.990
newstest2010.eng-fra	33.0	0.599	2489	66043	0.996
newstest2011.eng-fra	34.6	0.614	3003	80626	0.977
newstest2012.eng-fra	32.0	0.593	3003	78011	0.990
newstest2013.eng-fra	33.5	0.591	3000	70037	0.979
Tatoeba-test.eng-fra	50.8	0.671	10000	80769	0.981
tico19-test.eng-fra	41.8	0.638	2100	64655	1.000

opusTCv20210807+bt_transformer-big_2022-03-09.zip

• dataset: opusTCv20210807+bt
• model: transformer-big
• source language(s): eng
• target language(s): fra
• raw source language(s): eng
• raw target language(s): fra
• model: transformer-big
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt_transformer-big_2022-03-09.zip
• test set translations: opusTCv20210807+bt_transformer-big_2022-03-09.test.txt
• test set scores: opusTCv20210807+bt_transformer-big_2022-03-09.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdiscussdev2015-enfr.eng-fra	33.8	0.60628	1500	27986	1.000
newsdiscusstest2015-enfr.eng-fra	40.5	0.64981	1500	28027	0.995
newssyscomb2009.eng-fra	30.7	0.58905	502	12334	1.000
news-test2008.eng-fra	27.6	0.55516	2051	52685	1.000
newstest2009.eng-fra	30.0	0.57909	2525	69278	0.988
newstest2010.eng-fra	33.8	0.60293	2489	66043	0.996
newstest2011.eng-fra	35.0	0.61632	3003	80626	0.973
newstest2012.eng-fra	32.8	0.59737	3003	78011	0.985
newstest2013.eng-fra	34.6	0.59700	3000	70037	0.981
Tatoeba-test-v2021-08-07.eng-fra	53.1	0.69526	10000	84216	0.983
tico19-test.eng-fra	41.5	0.63470	2100	64655	0.974