opus2m-2020-08-01.yml

release: eng-gmq/opus2m-2020-08-01.zip
release-date: 2020-08-01
dataset-name: opus2m
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:
   - dan
   - fao
   - isl
   - nno
   - nob
   - non
   - swe
use-target-labels:
   - >>dan<<
   - >>fao<<
   - >>isl<<
   - >>nno<<
   - >>nob<<
   - >>nob_Hebr<<
   - >>non_Latn<<
   - >>swe<<
test-data:
   - Tatoeba-test.eng-dan: 10000/61469
   - Tatoeba-test.eng-fao: 294/1564
   - Tatoeba-test.eng-isl: 2500/16288
   - Tatoeba-test.eng-multi: 10000/60550
   - Tatoeba-test.eng-non: 15/109
   - Tatoeba-test.eng-nor: 5000/33820
   - Tatoeba-test.eng-swe: 10000/54832
BLEU-scores:
   - Tatoeba-test.eng-dan: 57.7
   - Tatoeba-test.eng-fao: 9.2
   - Tatoeba-test.eng-isl: 23.8
   - Tatoeba-test.eng-multi: 52.8
   - Tatoeba-test.eng-non: 0.7
   - Tatoeba-test.eng-nor: 50.3
   - Tatoeba-test.eng-swe: 57.8
chr-F-scores:
   - Tatoeba-test.eng-dan: 0.724
   - Tatoeba-test.eng-fao: 0.322
   - Tatoeba-test.eng-isl: 0.506
   - Tatoeba-test.eng-multi: 0.688
   - Tatoeba-test.eng-non: 0.196
   - Tatoeba-test.eng-nor: 0.678
   - Tatoeba-test.eng-swe: 0.717