opus-2021-02-23.yml

release: nor-fra/opus-2021-02-23.zip
release-date: 2021-02-23
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:
   - nno
   - nob
target-languages:
   - fra
training-data:
   - nno-fra: Tatoeba-train (175304) 
   - nob-fra: Tatoeba-train (8128830) 
validation-data:
   - fra-nno: Tatoeba-dev, 18
   - fra-nob: Tatoeba-dev, 982
   - total size of shuffled dev data: 998
   - devset = top 998  lines of Tatoeba-dev.src.shuffled!
test-data:
   - Tatoeba-test.nno-fra: 154/944
   - Tatoeba-test.nob-fra: 322/2261
   - Tatoeba-test.nor-fra: 476/3205
BLEU-scores:
   - Tatoeba-test.nno-fra: 54.8
   - Tatoeba-test.nob-fra: 53.4
   - Tatoeba-test.nor-fra: 55.2
chr-F-scores:
   - Tatoeba-test.nno-fra: 0.689
   - Tatoeba-test.nob-fra: 0.695
   - Tatoeba-test.nor-fra: 0.690