opus-2020-07-26.yml

release: eng-aav/opus-2020-07-26.zip
release-date: 2020-07-26
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:
   - eng
target-languages:
   - hoc
   - kha
   - khm
   - mnw
   - vie
use-target-labels:
   - >>hoc<<
   - >>hoc_Latn<<
   - >>kha<<
   - >>khm<<
   - >>khm_Latn<<
   - >>mnw<<
   - >>vie<<
   - >>vie_Hani<<
test-data:
   - Tatoeba-test.eng-hoc: 660/1894
   - Tatoeba-test.eng-kha: 1314/7819
   - Tatoeba-test.eng-khm: 752/1610
   - Tatoeba-test.eng-mnw: 9/40
   - Tatoeba-test.eng-multi: 5239/33106
   - Tatoeba-test.eng-vie: 2500/21591
BLEU-scores:
   - Tatoeba-test.eng-hoc: 0.1
   - Tatoeba-test.eng-kha: 0.4
   - Tatoeba-test.eng-khm: 0.2
   - Tatoeba-test.eng-mnw: 0.8
   - Tatoeba-test.eng-multi: 16.1
   - Tatoeba-test.eng-vie: 33.2
chr-F-scores:
   - Tatoeba-test.eng-hoc: 0.033
   - Tatoeba-test.eng-kha: 0.043
   - Tatoeba-test.eng-khm: 0.242
   - Tatoeba-test.eng-mnw: 0.003
   - Tatoeba-test.eng-multi: 0.311
   - Tatoeba-test.eng-vie: 0.508