zlw-eng

opus-2020-06-28.zip

• dataset: opus
• model: transformer
• source language(s): ces csb_Latn pol
• 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
newssyscomb2009-ceseng.ces.eng	23.8	0.521
newstest2009-ceseng.ces.eng	23.1	0.518
newstest2010-ceseng.ces.eng	23.3	0.527
newstest2011-ceseng.ces.eng	24.2	0.525
newstest2012-ceseng.ces.eng	23.4	0.521
newstest2013-ceseng.ces.eng	26.1	0.537
newstest2014-csen-ceseng.ces.eng	28.1	0.567
newstest2015-encs-ceseng.ces.eng	25.1	0.524
newstest2016-encs-ceseng.ces.eng	27.1	0.548
newstest2017-encs-ceseng.ces.eng	24.0	0.519
newstest2018-encs-ceseng.ces.eng	25.1	0.528
Tatoeba-test.ces-eng.ces.eng	50.8	0.672
Tatoeba-test.csb-eng.csb.eng	13.0	0.326
Tatoeba-test.multi.eng	49.4	0.662
Tatoeba-test.pol-eng.pol.eng	48.3	0.652

opus-2020-07-27.zip

• dataset: opus
• model: transformer
• source language(s): ces csb_Latn dsb hsb pol
• target language(s): eng
• model: transformer
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• 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
newssyscomb2009-ceseng.ces.eng	23.4	0.522
newstest2009-ceseng.ces.eng	22.9	0.515
newstest2010-ceseng.ces.eng	23.0	0.523
newstest2011-ceseng.ces.eng	23.9	0.523
newstest2012-ceseng.ces.eng	23.2	0.518
newstest2013-ceseng.ces.eng	25.9	0.535
newstest2014-csen-ceseng.ces.eng	27.7	0.563
newstest2015-encs-ceseng.ces.eng	25.4	0.525
newstest2016-encs-ceseng.ces.eng	26.8	0.545
newstest2017-encs-ceseng.ces.eng	23.9	0.519
newstest2018-encs-ceseng.ces.eng	24.7	0.526
Tatoeba-test.ces-eng.ces.eng	50.6	0.670
Tatoeba-test.csb-eng.csb.eng	7.0	0.279
Tatoeba-test.dsb-eng.dsb.eng	11.3	0.291
Tatoeba-test.hsb-eng.hsb.eng	20.2	0.415
Tatoeba-test.multi.eng	49.4	0.660
Tatoeba-test.pol-eng.pol.eng	48.1	0.649

opus2m-2020-08-01.zip

• dataset: opus2m
• model: transformer
• source language(s): ces csb_Latn dsb hsb pol
• 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
newssyscomb2009-ceseng.ces.eng	25.7	0.536
newstest2009-ceseng.ces.eng	24.6	0.530
newstest2010-ceseng.ces.eng	25.0	0.540
newstest2011-ceseng.ces.eng	25.9	0.539
newstest2012-ceseng.ces.eng	24.8	0.533
newstest2013-ceseng.ces.eng	27.8	0.551
newstest2014-csen-ceseng.ces.eng	30.3	0.585
newstest2015-encs-ceseng.ces.eng	27.5	0.542
newstest2016-encs-ceseng.ces.eng	29.1	0.564
newstest2017-encs-ceseng.ces.eng	26.0	0.537
newstest2018-encs-ceseng.ces.eng	27.3	0.544
Tatoeba-test.ces-eng.ces.eng	53.3	0.691
Tatoeba-test.csb-eng.csb.eng	10.2	0.313
Tatoeba-test.dsb-eng.dsb.eng	11.7	0.296
Tatoeba-test.hsb-eng.hsb.eng	24.6	0.426
Tatoeba-test.multi.eng	51.8	0.680
Tatoeba-test.pol-eng.pol.eng	50.4	0.667

opus4m-2020-08-12.zip

• dataset: opus4m
• model: transformer
• source language(s): ces csb_Latn dsb hsb pol
• target language(s): eng
• model: transformer
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus4m-2020-08-12.zip
• test set translations: opus4m-2020-08-12.test.txt
• test set scores: opus4m-2020-08-12.eval.txt

Benchmarks

testset	BLEU	chr-F
newssyscomb2009-ceseng.ces.eng	26.8	0.546
newstest2009-ceseng.ces.eng	26.1	0.541
newstest2010-ceseng.ces.eng	26.1	0.548
newstest2011-ceseng.ces.eng	27.1	0.548
newstest2012-ceseng.ces.eng	25.6	0.540
newstest2013-ceseng.ces.eng	29.0	0.560
newstest2014-csen-ceseng.ces.eng	31.8	0.596
newstest2015-encs-ceseng.ces.eng	29.0	0.553
newstest2016-encs-ceseng.ces.eng	30.1	0.572
newstest2017-encs-ceseng.ces.eng	27.4	0.546
newstest2018-encs-ceseng.ces.eng	28.5	0.553
Tatoeba-test.ces-eng.ces.eng	54.4	0.699
Tatoeba-test.csb-eng.csb.eng	15.6	0.332
Tatoeba-test.dsb-eng.dsb.eng	20.0	0.360
Tatoeba-test.hsb-eng.hsb.eng	15.6	0.382
Tatoeba-test.multi.eng	52.8	0.687
Tatoeba-test.pol-eng.pol.eng	51.1	0.674

opus1m+bt-2021-05-02.zip

• dataset: opus1m+bt
• model: transformer-align
• source language(s): ces csb dsb hsb pol
• target language(s): eng
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus1m+bt-2021-05-02.zip
• test set translations: opus1m+bt-2021-05-02.test.txt
• test set scores: opus1m+bt-2021-05-02.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newssyscomb2009.ces-eng	23.0	0.516	502	11821	0.980
newstest2009.ces-eng	22.4	0.511	2525	65402	0.966
newstest2010.ces-eng	22.4	0.518	2489	61724	0.983
newstest2011.ces-eng	23.8	0.521	3003	74681	0.995
newstest2012.ces-eng	22.5	0.513	3003	72812	1.000
newstest2013.ces-eng	25.2	0.529	3000	64505	1.000
newstest2014-csen.ces-eng	26.9	0.556	3003	68065	1.000
newstest2015-encs.ces-eng	24.5	0.517	2656	53572	0.972
newstest2016-encs.ces-eng	26.6	0.544	2999	64670	1.000
newstest2017-encs.ces-eng	23.3	0.514	3005	61725	0.999
newstest2018-encs.ces-eng	24.3	0.522	2983	63496	0.993
Tatoeba-test.ces-eng	49.8	0.663	10000	75370	0.953
Tatoeba-test.csb-eng	6.0	0.271	27	237	1.000
Tatoeba-test.dsb-eng	11.1	0.286	34	214	1.000
Tatoeba-test.hsb-eng	18.2	0.400	40	246	1.000
Tatoeba-test.multi-eng	48.6	0.652	10000	75346	0.961
Tatoeba-test.pol-eng	47.6	0.645	10000	74996	0.962

opus4m+btTCv20210807-2021-10-01.zip

• dataset: opus4m+btTCv20210807
• model: transformer
• source language(s): ces csb dsb hsb pol slk szl
• target language(s): eng
• 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:
• download: opus4m+btTCv20210807-2021-10-01.zip
• test set translations: opus4m+btTCv20210807-2021-10-01.test.txt
• test set scores: opus4m+btTCv20210807-2021-10-01.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newssyscomb2009.ces-eng	27.3	0.544	502	11821	0.973
newstest2009.ces-eng	25.5	0.536	2525	65402	0.972
newstest2010.ces-eng	26.6	0.550	2489	61724	0.981
newstest2011.ces-eng	26.8	0.547	3003	74681	0.993
newstest2012.ces-eng	25.5	0.540	3003	72812	1.000
newstest2013.ces-eng	28.8	0.561	3000	64505	1.000
newstest2014-csen.ces-eng	32.1	0.597	3003	68065	1.000
newstest2015-encs.ces-eng	28.6	0.552	2656	53572	0.973
newstest2016-encs.ces-eng	30.8	0.575	2999	64670	0.998
newstest2017-encs.ces-eng	27.0	0.543	3005	61725	0.994
newstest2018-encs.ces-eng	28.4	0.555	2983	63496	0.988
Tatoeba-test-v2021-08-07.multi-eng	50.4	0.666	10000	75475	0.963
Tatoeba-test-v2021-08-07.multi-multi	50.4	0.666	10000	75475	0.963

opusTCv20210807+bt_transformer-big_2022-03-17.zip

• dataset: opusTCv20210807+bt
• model: transformer-big
• source language(s): ces csb csb_Latn dsb hsb pol slk szl
• target language(s): eng
• raw source language(s): ces csb dsb hsb pol slk szl
• raw target language(s): eng
• model: transformer-big
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt_transformer-big_2022-03-17.zip
• test set translations: opusTCv20210807+bt_transformer-big_2022-03-17.test.txt
• test set scores: opusTCv20210807+bt_transformer-big_2022-03-17.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newssyscomb2009.ces-eng	30.3	0.56499	502	11821	0.976
newstest2009.ces-eng	29.5	0.56308	2525	65402	0.969
newstest2010.ces-eng	30.7	0.57791	2489	61724	0.974
newstest2011.ces-eng	30.9	0.57336	3003	74681	0.988
newstest2012.ces-eng	29.4	0.56767	3003	72812	1.000
newstest2013.ces-eng	32.8	0.58809	3000	64505	1.000
newstest2014-csen.ces-eng	38.8	0.64427	3003	68065	0.996
newstest2015-encs.ces-eng	33.4	0.58607	2656	53572	0.968
newstest2016-encs.ces-eng	37.1	0.61780	2999	64670	0.992
newstest2017-encs.ces-eng	32.5	0.58257	3005	61725	0.988
newstest2018-encs.ces-eng	33.1	0.58678	2983	63496	0.982
Tatoeba-test-v2021-08-07.ces-eng	57.2	0.71663	13824	104993	0.961
Tatoeba-test-v2021-08-07.csb-eng	19.1	0.38897	27	237	1.000
Tatoeba-test-v2021-08-07.dsb-eng	22.7	0.37153	105	567	0.949
Tatoeba-test-v2021-08-07.hsb-eng	45.1	0.51040	154	919	0.961
Tatoeba-test-v2021-08-07.multi-eng	56.4	0.70984	10000	75698	0.962
Tatoeba-test-v2021-08-07.pol-eng	55.6	0.70499	10099	75760	0.966