opus-2020-07-06.yml

release: eng-zls/opus-2020-07-06.zip
release-date: 2020-07-06
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:
   - bos
   - bul
   - hrv
   - mkd
   - slv
   - srp
use-target-labels:
   - >>bos_Latn<<
   - >>bul<<
   - >>bul_Latn<<
   - >>hrv<<
   - >>mkd<<
   - >>slv<<
   - >>srp_Cyrl<<
   - >>srp_Latn<<
test-data:
   - Tatoeba-test.eng-bul: 10000/56740
   - Tatoeba-test.eng-hbs: 10000/52571
   - Tatoeba-test.eng-mkd: 10000/51245
   - Tatoeba-test.eng-multi: 10000/53384
   - Tatoeba-test.eng-slv: 2007/9410
BLEU-scores:
   - Tatoeba-test.eng-bul: 46.2
   - Tatoeba-test.eng-hbs: 0.8
   - Tatoeba-test.eng-mkd: 43.5
   - Tatoeba-test.eng-multi: 42.4
   - Tatoeba-test.eng-slv: 17.6
chr-F-scores:
   - Tatoeba-test.eng-bul: 0.646
   - Tatoeba-test.eng-hbs: 0.051
   - Tatoeba-test.eng-mkd: 0.629
   - Tatoeba-test.eng-multi: 0.612
   - Tatoeba-test.eng-slv: 0.348