opus-2020-08-12.yml

release: mkh-eng/opus-2020-08-12.zip
release-date: 2020-08-12
dataset-name: opus
modeltype: transformer
pre-processing: normalization + SentencePiece (spm32k,spm32k)
subwords:
   source: spm32k
   target: spm32k
subword-models:
   source: source.spm
   target: target.spm
source-languages:
   - kha
   - khm
   - mnw
   - vie
target-languages:
   - eng
test-data:
   Tatoeba-test.kha-eng: 1314/6985
   Tatoeba-test.khm-eng: 752/3505
   Tatoeba-test.mnw-eng: 9/67
   Tatoeba-test.multi-eng: 4579/28701
   Tatoeba-test.vie-eng: 2500/17313
BLEU-scores:
   Tatoeba-test.kha-eng: 0.5
   Tatoeba-test.khm-eng: 8.5
   Tatoeba-test.mnw-eng: 0.7
   Tatoeba-test.multi-eng: 24.5
   Tatoeba-test.vie-eng: 34.4
chr-F-scores:
   Tatoeba-test.kha-eng: 0.108
   Tatoeba-test.khm-eng: 0.206
   Tatoeba-test.mnw-eng: 0.110
   Tatoeba-test.multi-eng: 0.407
   Tatoeba-test.vie-eng: 0.529