Neurora/opus-tatoeba-eng-ces

Opus Tatoeba | English -> Czech

• dataset: opus
• model: transformer
• source language(s): eng
• target language(s): ces
• model: transformer
• 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
newssyscomb2009.eng-ces	23.4	0.513	502	10032	0.992
news-test2008.eng-ces	20.7	0.488	2051	42484	0.993
newstest2009.eng-ces	21.9	0.502	2525	55533	0.992
newstest2010.eng-ces	22.1	0.507	2489	52958	0.996
newstest2011.eng-ces	23.3	0.511	3003	65653	0.966
newstest2012.eng-ces	20.7	0.483	3003	65456	0.951
newstest2013.eng-ces	24.9	0.518	3000	57250	0.970
newstest2015-encs.eng-ces	24.4	0.527	2656	45931	1.000
newstest2016-encs.eng-ces	27.5	0.546	2999	57013	1.000
newstest2017-encs.eng-ces	22.7	0.506	3005	54461	0.986
newstest2018-encs.eng-ces	22.4	0.506	2983	54772	1.000
newstest2019-encs.eng-ces	23.8	0.517	1997	43373	0.992
Tatoeba-test.eng-ces	47.8	0.663	10000	65288	0.988