fin-eng

opus-2021-02-18.zip

• dataset: opus
• model: transformer-align
• source language(s): fin
• target language(s): eng
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus-2021-02-18.zip
• test set translations: opus-2021-02-18.test.txt
• test set scores: opus-2021-02-18.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.fin-eng	25.3	0.536	1500	32104	1.000
newstest2015-enfi.fin-eng	26.9	0.547	1370	27356	0.997
newstest2016-enfi.fin-eng	29.0	0.571	3000	63043	1.000
newstest2017-enfi.fin-eng	32.3	0.594	3002	61936	0.997
newstest2018-enfi.fin-eng	23.8	0.517	3000	62325	0.991
newstest2019-fien.fin-eng	29.0	0.565	1996	36227	1.000
newstestB2016-enfi.fin-eng	24.5	0.527	3000	63043	0.999
newstestB2017-enfi.fin-eng	27.4	0.557	3002	61936	1.000
newstestB2017-fien.fin-eng	27.4	0.557	3002	61936	1.000
Tatoeba-test.fin-eng	53.4	0.697	10000	74651	0.990

opus+bt-2021-04-30.zip

• dataset: opus+bt
• model: transformer-align
• source language(s): fin
• target language(s): eng
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus+bt-2021-04-30.zip
• test set translations: opus+bt-2021-04-30.test.txt
• test set scores: opus+bt-2021-04-30.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.fin-eng	25.8	0.541	1500	32104	1.000
newstest2015-enfi.fin-eng	27.1	0.551	1370	27356	0.989
newstest2016-enfi.fin-eng	30.2	0.577	3000	63043	1.000
newstest2017-enfi.fin-eng	33.3	0.602	3002	61936	0.992
newstest2018-enfi.fin-eng	24.1	0.521	3000	62325	0.989
newstest2019-fien.fin-eng	29.7	0.570	1996	36227	0.999
newstestB2016-enfi.fin-eng	24.9	0.531	3000	63043	0.992
newstestB2017-enfi.fin-eng	28.1	0.563	3002	61936	1.000
newstestB2017-fien.fin-eng	28.1	0.563	3002	61936	1.000
Tatoeba-test.fin-eng	53.1	0.695	10000	74651	0.988

opusTCv20210807+bt-2021-08-25.zip

• dataset: opusTCv20210807+bt
• model: transformer-align
• source language(s): fin
• target language(s): eng
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt-2021-08-25.zip
• test set translations: opusTCv20210807+bt-2021-08-25.test.txt
• test set scores: opusTCv20210807+bt-2021-08-25.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.fin-eng	27.1	0.550	1500	32104	0.988
newstest2015-enfi.fin-eng	28.5	0.560	1370	27356	0.980
newstest2016-enfi.fin-eng	31.7	0.586	3000	63043	1.000
newstest2017-enfi.fin-eng	34.6	0.610	3002	61936	0.988
newstest2018-enfi.fin-eng	25.4	0.530	3000	62325	0.981
newstest2019-fien.fin-eng	30.6	0.577	1996	36227	0.994
newstestB2016-enfi.fin-eng	25.8	0.538	3000	63043	0.987
newstestB2017-enfi.fin-eng	29.6	0.572	3002	61936	0.999
newstestB2017-fien.fin-eng	29.6	0.572	3002	61936	0.999
Tatoeba-test-v2021-08-07.fin-eng	54.1	0.700	10000	75212	0.988

opusTCv20210807+bt-2021-11-09.zip

• dataset: opusTCv20210807+bt
• model: transformer-big-align
• source language(s): fin
• target language(s): eng
• raw source language(s): fin
• raw target language(s): eng
• model: transformer-big-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+bt-2021-11-09.zip
• test set translations: opusTCv20210807+bt-2021-11-09.test.txt
• test set scores: opusTCv20210807+bt-2021-11-09.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.fin-eng	28.2	0.559	1500	32104	0.993
newstest2015-enfi.fin-eng	29.4	0.568	1370	27356	0.979
newstest2016-enfi.fin-eng	33.0	0.597	3000	63043	1.000
newstest2017-enfi.fin-eng	36.5	0.623	3002	61936	0.992
newstest2018-enfi.fin-eng	26.2	0.536	3000	62325	0.983
newstest2019-fien.fin-eng	31.6	0.584	1996	36227	0.992
newstestB2016-enfi.fin-eng	27.2	0.548	3000	63043	0.989
newstestB2017-enfi.fin-eng	30.9	0.582	3002	61936	1.000
newstestB2017-fien.fin-eng	30.9	0.582	3002	61936	1.000
Tatoeba-test-v2021-08-07.fin-eng	55.6	0.710	10000	75212	0.992

opusTCv20210807+bt-2021-12-08.zip

• dataset: opusTCv20210807+bt
• model: transformer-big
• source language(s): fin
• target language(s): eng
• raw source language(s): fin
• raw target language(s): eng
• 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.fin-eng	29.0	0.5647	1500	32104	0.994
newstest2015-enfi.fin-eng	30.4	0.5771	1370	27356	0.983
newstest2016-enfi.fin-eng	34.7	0.6090	3000	63043	1.000
newstest2017-enfi.fin-eng	37.6	0.6318	3002	61936	0.993
newstest2018-enfi.fin-eng	27.1	0.5420	3000	62325	0.987
newstest2019-fien.fin-eng	33.2	0.5991	1996	36227	1.000
newstestB2016-enfi.fin-eng	28.4	0.5571	3000	63043	0.997
newstestB2017-enfi.fin-eng	31.6	0.5903	3002	61936	1.000
newstestB2017-fien.fin-eng	31.6	0.5903	3002	61936	1.000
Tatoeba-test-v2021-08-07.fin-eng	57.3	0.7232	10000	75212	0.994

opusTCv20210807+nopar+ft95-sepvoc_transformer-align_2022-01-28.zip

• dataset: opusTCv20210807+nopar+ft95
• model: transformer-align
• source language(s): fin
• target language(s): eng
• raw source language(s): fin
• raw target language(s): eng
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+nopar+ft95-sepvoc_transformer-align_2022-01-28.zip
• test set translations: opusTCv20210807+nopar+ft95-sepvoc_transformer-align_2022-01-28.test.txt
• test set scores: opusTCv20210807+nopar+ft95-sepvoc_transformer-align_2022-01-28.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.fin-eng	27.2	0.55238	1500	32104	0.983
newstest2015-enfi.fin-eng	29.0	0.56510	1370	27356	0.975
newstest2016-enfi.fin-eng	32.4	0.59189	3000	63043	0.999
newstest2017-enfi.fin-eng	35.3	0.61651	3002	61936	0.980
newstest2018-enfi.fin-eng	25.3	0.52998	3000	62325	0.977
newstest2019-fien.fin-eng	31.2	0.58389	1996	36227	0.987
newstestB2016-enfi.fin-eng	26.5	0.54374	3000	63043	0.980
newstestB2017-enfi.fin-eng	29.8	0.57630	3002	61936	0.993
newstestB2017-fien.fin-eng	29.8	0.57630	3002	61936	0.993
Tatoeba-test-v2021-08-07.fin-eng	54.5	0.70539	10000	75524	0.984
Tatoeba-test-v2021-08-07.intgemm8.fin-eng	54.1	0.70224	10000	75524	0.983
Tatoeba-test-v2021-08-07.intgemm8shiftAlphas.fin-eng	54.2	0.70341	10000	75524	0.984
Tatoeba-test-v2021-08-07.intgemm8shiftAlphas.shortlist.fin-eng	54.1	0.70286	10000	75524	0.984
Tatoeba-test-v2021-08-07.intgemm8shift.fin-eng	54.2	0.70275	10000	75524	0.984
Tatoeba-test-v2021-08-07.intgemm8shift.shortlist.fin-eng	54.1	0.70241	10000	75524	0.984
Tatoeba-test-v2021-08-07.intgemm8.shortlist.fin-eng	54.0	0.70140	10000	75524	0.983

opusTCv20210807+news+bt_transformer-big_2023-04-13.zip

• dataset: opusTCv20210807+news+bt
• model: transformer-big
• source language(s): fin
• target language(s): eng
• raw source language(s): fin
• raw target language(s): eng
• model: transformer-big
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opusTCv20210807+news+bt_transformer-big_2023-04-13.zip
• test set translations: opusTCv20210807+news+bt_transformer-big_2023-04-13.test.txt
• test set scores: opusTCv20210807+news+bt_transformer-big_2023-04-13.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2015-enfi.fin-eng	29.1	0.56442	1500	32104	0.983
newstest2015-enfi.fin-eng	30.3	0.57413	1370	27356	0.976
newstest2016-enfi.fin-eng	36.3	0.61780	3000	63043	1.000
newstest2017-enfi.fin-eng	38.4	0.63566	3002	61936	0.982
newstest2018-enfi.fin-eng	26.9	0.54142	3000	62325	0.981
newstest2019-fien.fin-eng	30.9	0.57880	1996	36227	0.987
newstestB2016-enfi.fin-eng	28.7	0.55994	3000	63043	0.986
newstestB2017-enfi.fin-eng	32.3	0.59170	3002	61936	0.996
Tatoeba-test-v2021-08-07.fin-eng	56.4	0.71708	10000	75524	0.982