opus-2020-06-28.yml

release: eng-phi/opus-2020-06-28.zip
release-date: 2020-06-28
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:
   - akl
   - ceb
   - hil
   - ilo
   - pag
   - pmn
   - war
use-target-labels:
   - >>akl_Latn<<
   - >>ceb<<
   - >>hil<<
   - >>ilo<<
   - >>pag<<
   - >>pmn<<
   - >>war<<
test-data:
   - Tatoeba-test.eng-akl: 27/69
   - Tatoeba-test.eng-ceb: 378/1678
   - Tatoeba-test.eng-hil: 22/104
   - Tatoeba-test.eng-ilo: 1093/6040
   - Tatoeba-test.eng-multi: 0/0
   - Tatoeba-test.eng-pag: 49/268
   - Tatoeba-test.eng-pmn: 1497/7499
   - Tatoeba-test.eng-war: 1512/9364
BLEU-scores:
   - Tatoeba-test.eng-akl: 6.2
   - Tatoeba-test.eng-ceb: 10.6
   - Tatoeba-test.eng-hil: 17.1
   - Tatoeba-test.eng-ilo: 33.9
   - Tatoeba-test.eng-multi: 13.6
   - Tatoeba-test.eng-pag: 16.8
   - Tatoeba-test.eng-pmn: 0.5
   - Tatoeba-test.eng-war: 12.8
chr-F-scores:
   - Tatoeba-test.eng-akl: 0.245
   - Tatoeba-test.eng-ceb: 0.436
   - Tatoeba-test.eng-hil: 0.490
   - Tatoeba-test.eng-ilo: 0.587
   - Tatoeba-test.eng-multi: 0.392
   - Tatoeba-test.eng-pag: 0.484
   - Tatoeba-test.eng-pmn: 0.163
   - Tatoeba-test.eng-war: 0.437