eng-fin

opus-2021-02-19.zip

• dataset: opus
• model: transformer-align
• source language(s): eng
• target language(s): fin
• 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
newsdev2015-enfi.eng-fin	21.6	0.556	1500	23375	1.000
newstest2015-enfi.eng-fin	23.2	0.567	1370	19968	1.000
newstest2016-enfi.eng-fin	24.9	0.578	3000	48116	0.986
newstest2017-enfi.eng-fin	27.5	0.605	3002	45718	0.996
newstest2018-enfi.eng-fin	18.4	0.532	3000	45475	1.000
newstest2019-enfi.eng-fin	23.3	0.551	1997	38369	0.966
newstestB2016-enfi.eng-fin	19.7	0.542	3000	45766	1.000
newstestB2017-enfi.eng-fin	22.7	0.565	3002	45506	1.000
Tatoeba-test.eng-fin	38.7	0.629	10000	60517	0.935

opus+bt-2021-03-09.zip

• dataset: opus+bt
• model: transformer-align
• source language(s): eng
• target language(s): fin
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus+bt-2021-03-09.zip
• test set translations: opus+bt-2021-03-09.test.txt
• test set scores: opus+bt-2021-03-09.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.eng-fin	22.9	0.569	1500	23375	1.000
newstest2015-enfi.eng-fin	25.0	0.586	1370	19968	1.000
newstest2016-enfi.eng-fin	26.3	0.595	3000	48116	0.981
newstest2017-enfi.eng-fin	29.3	0.624	3002	45718	0.982
newstest2018-enfi.eng-fin	19.7	0.545	3000	45475	1.000
newstest2019-enfi.eng-fin	24.1	0.565	1997	38369	0.957
newstestB2016-enfi.eng-fin	21.3	0.558	3000	45766	1.000
newstestB2017-enfi.eng-fin	24.5	0.583	3002	45506	0.987
Tatoeba-test.eng-fin	38.7	0.633	10000	60517	0.936

opusTCv20210807+bt-2021-09-01.zip

• dataset: opusTCv20210807+bt
• model: transformer-align
• source language(s): eng
• target language(s): fin
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt-2021-09-01.zip
• test set translations: opusTCv20210807+bt-2021-09-01.test.txt
• test set scores: opusTCv20210807+bt-2021-09-01.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.eng-fin	24.3	0.576	1500	23375	1.000
newstest2015-enfi.eng-fin	26.0	0.592	1370	19968	1.000
newstest2016-enfi.eng-fin	28.0	0.605	3000	48116	0.993
newstest2017-enfi.eng-fin	31.1	0.634	3002	45718	0.996
newstest2018-enfi.eng-fin	20.6	0.550	3000	45475	1.000
newstest2019-enfi.eng-fin	26.2	0.578	1997	38369	0.973
newstestB2016-enfi.eng-fin	22.1	0.565	3000	45766	1.000
newstestB2017-enfi.eng-fin	25.7	0.591	3002	45506	1.000
Tatoeba-test-v2021-08-07.eng-fin	39.7	0.642	10000	60797	0.948

opusTCv20210807+bt-2021-11-08.zip

• dataset: opusTCv20210807+bt
• model: transformer-big-align
• source language(s): eng
• target language(s): fin
• raw source language(s): eng
• raw target language(s): fin
• model: transformer-big-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt-2021-11-08.zip
• test set translations: opusTCv20210807+bt-2021-11-08.test.txt
• test set scores: opusTCv20210807+bt-2021-11-08.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.eng-fin	24.6	0.580	1500	23375	1.000
newstest2015-enfi.eng-fin	26.5	0.595	1370	19968	1.000
newstest2016-enfi.eng-fin	28.2	0.607	3000	48116	0.993
newstest2017-enfi.eng-fin	31.4	0.636	3002	45718	0.992
newstest2018-enfi.eng-fin	20.7	0.553	3000	45475	1.000
newstest2019-enfi.eng-fin	26.5	0.581	1997	38369	0.965
newstestB2016-enfi.eng-fin	22.3	0.568	3000	45766	1.000
Tatoeba-test-v2021-08-07.eng-fin	39.9	0.643	10000	60797	0.949

opusTCv20210807+bt-2021-12-08.zip

• dataset: opusTCv20210807+bt
• model: transformer-big
• source language(s): eng
• target language(s): fin
• raw source language(s): eng
• raw target language(s): fin
• model: transformer-big
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt-2021-12-08.zip
• test set translations: opusTCv20210807+bt-2021-12-08.test.txt
• test set scores: opusTCv20210807+bt-2021-12-08.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.eng-fin	25.3	0.5846	1500	23375	1.000
newstest2015-enfi.eng-fin	27.9	0.6039	1370	19968	1.000
newstest2016-enfi.eng-fin	29.6	0.6180	3000	48116	0.991
newstest2017-enfi.eng-fin	32.7	0.6486	3002	45718	0.993
newstest2018-enfi.eng-fin	21.5	0.5602	3000	45475	1.000
newstest2019-enfi.eng-fin	26.9	0.5863	1997	38369	0.968
newstestB2016-enfi.eng-fin	23.4	0.5758	3000	45766	1.000
newstestB2017-enfi.eng-fin	27.3	0.6051	3002	45506	0.998
Tatoeba-test-v2021-08-07.eng-fin	38.9	0.6385	10000	60797	0.941