phi-phi

opus-2020-10-04.zip

• dataset: opus
• model: transformer
• source language(s): akl_Latn ceb eng hil ilo pag war
• target language(s): akl_Latn ceb eng hil ilo pag war
• 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)
• download: opus-2020-10-04.zip
• test set translations: opus-2020-10-04.test.txt
• test set scores: opus-2020-10-04.eval.txt

Benchmarks

testset	BLEU	chr-F
Tatoeba-test.akl-eng.akl.eng	9.4	0.287
Tatoeba-test.ceb-eng.ceb.eng	21.7	0.404
Tatoeba-test.eng-akl.eng.akl	3.2	0.188
Tatoeba-test.eng-ceb.eng.ceb	10.8	0.432
Tatoeba-test.eng-hil.eng.hil	17.9	0.517
Tatoeba-test.eng-ilo.eng.ilo	33.7	0.588
Tatoeba-test.eng-pag.eng.pag	16.2	0.488
Tatoeba-test.eng-war.eng.war	12.8	0.435
Tatoeba-test.hil-eng.hil.eng	17.7	0.385
Tatoeba-test.ilo-eng.ilo.eng	37.0	0.559
Tatoeba-test.multi.multi	21.7	0.465
Tatoeba-test.pag-eng.pag.eng	21.3	0.465
Tatoeba-test.war-eng.war.eng	16.1	0.365