art-por

opus1m-2021-05-15.zip

• dataset: opus1m
• model: transformer-align
• source language(s): avk bzt epo ido ile ina jbo lfn nov
• target language(s): pob por
• model: transformer-align
• 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: >>pob<< >>por<<
• download: opus1m-2021-05-15.zip
• test set translations: opus1m-2021-05-15.test.txt
• test set scores: opus1m-2021-05-15.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
Tatoeba-test.avk-por	0.0	0.061	1	6	0.368
Tatoeba-test.bzt-por	1.5	0.098	14	71	1.000
Tatoeba-test.epo-por	17.6	0.393	10000	89903	0.940
Tatoeba-test.ido-por	5.4	0.256	20	98	1.000
Tatoeba-test.ile-por	1.6	0.160	57	335	1.000
Tatoeba-test.ina-por	2.4	0.208	2500	23444	1.000
Tatoeba-test.jbo-por	0.0	0.078	2	9	0.287
Tatoeba-test.lfn_Cyrl-por	0.2	0.114	220	1927	1.000
Tatoeba-test.lfn_Latn-por	1.7	0.182	1725	22989	0.993
Tatoeba-test.lfn-por	1.5	0.176	1945	24916	1.000
Tatoeba-test.multi-por	12.3	0.318	10000	95398	0.998
Tatoeba-test.nov-por	5.5	0.151	3	15	1.000
Tatoeba-test.sjn-por	0.0	0.070	2	9	0.449
Tatoeba-test.tlh-por	1.0	0.080	22	125	0.456
Tatoeba-test.tzl-por	1.2	0.121	15	85	0.779
Tatoeba-test.vol-por	1.5	0.111	15	80	0.748