opus-2020-06-16.zip dataset: opus model: transformer-align source language(s): bel bel_Latn rus ukr target language(s): bel bel_Latn rus ukr 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) download: opus-2020-06-16.zip test set translations: opus-2020-06-16.test.txt test set scores: opus-2020-06-16.eval.txt Benchmarks testset BLEU chr-F Tatoeba-test.bel-rus.bel.rus 54.4 0.738 Tatoeba-test.bel+rus+ukr.bel+rus+ukr 59.4 0.759 Tatoeba-test.bel.ukr 59.5 0.760 Tatoeba-test.bel-ukr.bel.ukr 53.7 0.730 Tatoeba-test.rus-bel.rus.bel 48.9 0.696 Tatoeba-test.rus-ukr.rus.ukr 60.7 0.768 Tatoeba-test.ukr-bel.ukr.bel 49.9 0.704 Tatoeba-test.ukr-rus.ukr.rus 65.7 0.800