bat-eng

opus-2020-06-28.zip

• dataset: opus
• model: transformer
• source language(s): lav lit ltg prg_Latn sgs
• 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
newsdev2017-enlv-laveng.lav.eng	25.4	0.551
newsdev2019-enlt-liteng.lit.eng	25.3	0.537
newstest2017-enlv-laveng.lav.eng	20.0	0.498
newstest2019-lten-liteng.lit.eng	28.9	0.576
Tatoeba-test.lav-eng.lav.eng	48.8	0.666
Tatoeba-test.lit-eng.lit.eng	52.9	0.683
Tatoeba-test.multi.eng	48.5	0.645
Tatoeba-test.prg-eng.prg.eng	1.4	0.160
Tatoeba-test.sgs-eng.sgs.eng	22.4	0.334

opus-2020-07-26.zip

• dataset: opus
• model: transformer
• source language(s): lav lit ltg prg_Latn sgs
• target language(s): eng
• model: transformer
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus-2020-07-26.zip
• test set translations: opus-2020-07-26.test.txt
• test set scores: opus-2020-07-26.eval.txt

Benchmarks

testset	BLEU	chr-F
newsdev2017-enlv-laveng.lav.eng	25.9	0.553
newsdev2019-enlt-liteng.lit.eng	24.9	0.535
newstest2017-enlv-laveng.lav.eng	19.5	0.496
newstest2019-lten-liteng.lit.eng	28.0	0.575
Tatoeba-test.lav-eng.lav.eng	48.7	0.662
Tatoeba-test.lit-eng.lit.eng	52.6	0.684
Tatoeba-test.multi.eng	48.2	0.643
Tatoeba-test.prg-eng.prg.eng	0.8	0.155
Tatoeba-test.sgs-eng.sgs.eng	13.4	0.325

opus2m-2020-07-31.zip

• dataset: opus2m
• model: transformer
• source language(s): lav lit ltg prg_Latn sgs
• target language(s): eng
• model: transformer
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus2m-2020-07-31.zip
• test set translations: opus2m-2020-07-31.test.txt
• test set scores: opus2m-2020-07-31.eval.txt

Benchmarks

testset	BLEU	chr-F
newsdev2017-enlv-laveng.lav.eng	27.5	0.566
newsdev2019-enlt-liteng.lit.eng	27.8	0.557
newstest2017-enlv-laveng.lav.eng	21.1	0.512
newstest2019-lten-liteng.lit.eng	30.2	0.592
Tatoeba-test.lav-eng.lav.eng	51.5	0.687
Tatoeba-test.lit-eng.lit.eng	55.1	0.703
Tatoeba-test.multi.eng	50.6	0.662
Tatoeba-test.prg-eng.prg.eng	1.0	0.159
Tatoeba-test.sgs-eng.sgs.eng	16.5	0.265

opus4m-2020-08-12.zip

• dataset: opus4m
• model: transformer
• source language(s): lav lit ltg prg_Latn sgs
• 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
newsdev2017-enlv-laveng.lav.eng	28.3	0.574
newsdev2019-enlt-liteng.lit.eng	28.8	0.563
newstest2017-enlv-laveng.lav.eng	22.1	0.517
newstest2019-lten-liteng.lit.eng	31.3	0.602
Tatoeba-test.lav-eng.lav.eng	52.3	0.692
Tatoeba-test.lit-eng.lit.eng	55.9	0.708
Tatoeba-test.multi.eng	51.2	0.666
Tatoeba-test.prg-eng.prg.eng	0.8	0.153
Tatoeba-test.sgs-eng.sgs.eng	14.1	0.303

opus1m+bt-2021-05-01.zip

• dataset: opus1m+bt
• model: transformer-align
• source language(s): lav lit ltg prg sgs
• target language(s): eng
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus1m+bt-2021-05-01.zip
• test set translations: opus1m+bt-2021-05-01.test.txt
• test set scores: opus1m+bt-2021-05-01.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2017-enlv.lav-eng	25.8	0.554	2003	48175	1.000
newsdev2019-enlt.lit-eng	25.1	0.534	2000	49666	0.980
newstest2017-enlv.lav-eng	19.8	0.498	2001	47511	1.000
newstest2019-lten.lit-eng	27.7	0.573	1000	26079	0.956
Tatoeba-test.lav-eng	49.7	0.669	1631	11212	0.979
Tatoeba-test.lit-eng	50.0	0.662	2500	17686	0.971
Tatoeba-test.ltg-eng	18.0	0.328	1	5	1.000
Tatoeba-test.multi-eng	48.2	0.641	4396	30772	0.980
Tatoeba-test.prg-eng	0.7	0.157	213	1663	1.000
Tatoeba-test.sgs-eng	16.9	0.294	52	207	1.000

opus4m+btTCv20210807-2021-09-30.zip

• dataset: opus4m+btTCv20210807
• model: transformer
• source language(s): lav lit ltg prg sgs
• target language(s): eng
• model: transformer
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus4m+btTCv20210807-2021-09-30.zip
• test set translations: opus4m+btTCv20210807-2021-09-30.test.txt
• test set scores: opus4m+btTCv20210807-2021-09-30.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
newsdev2017-enlv.lav-eng	26.5	0.515	2003	48175	0.999
newsdev2019-enlt.lit-eng	30.1	0.575	2000	49666	0.950
newstest2017-enlv.lav-eng	19.5	0.457	2001	47511	0.972
newstest2019-lten.lit-eng	31.1	0.599	1000	26079	0.933
Tatoeba-test-v2021-08-07.lav-eng	53.9	0.700	1631	11212	0.997
Tatoeba-test-v2021-08-07.lit-eng	55.7	0.706	2528	17853	0.982
Tatoeba-test-v2021-08-07.multi-eng	52.7	0.675	4424	30939	0.993
Tatoeba-test-v2021-08-07.multi-multi	52.7	0.675	4424	30939	0.993
Tatoeba-test-v2021-08-07.prg-eng	1.3	0.158	213	1663	1.000
Tatoeba-test-v2021-08-07.sgs-eng	17.2	0.256	52	207	0.985