opus-2020-10-04.yml

release: cpp-cpp/opus-2020-10-04.zip
release-date: 2020-10-04
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
   - ind
   - jak
   - max
   - min
   - msa
   - pap
   - tmw
   - zlm
   - zsm
target-languages:
   - eng
   - ind
   - jak
   - max
   - min
   - msa
   - pap
   - tmw
   - zlm
   - zsm
use-target-labels:
   - ">>eng<<"
   - ">>ind<<"
   - ">>jak_Latn<<"
   - ">>max_Latn<<"
   - ">>min<<"
   - ">>msa_Latn<<"
   - ">>pap<<"
   - ">>tmw_Latn<<"
   - ">>zlm<<"
   - ">>zlm_Latn<<"
   - ">>zsm_Latn<<"
test-data:
   Tatoeba-test.eng-msa: 5000/27896
   Tatoeba-test.eng-pap: 70/287
   Tatoeba-test.msa-eng: 5000/31183
   Tatoeba-test.msa-msa: 870/4915
   Tatoeba-test.msa-pap: 2/8
   Tatoeba-test.multi-multi: 10000/58802
   Tatoeba-test.pap-eng: 70/275
   Tatoeba-test.pap-msa: 2/6
BLEU-scores:
   Tatoeba-test.eng-msa: 33.2
   Tatoeba-test.eng-pap: 45.1
   Tatoeba-test.msa-eng: 39.8
   Tatoeba-test.msa-msa: 14.8
   Tatoeba-test.msa-pap: 31.7
   Tatoeba-test.multi-multi: 37.0
   Tatoeba-test.pap-eng: 48.9
   Tatoeba-test.pap-msa: 17.7
chr-F-scores:
   Tatoeba-test.eng-msa: 0.587
   Tatoeba-test.eng-pap: 0.638
   Tatoeba-test.msa-eng: 0.585
   Tatoeba-test.msa-msa: 0.353
   Tatoeba-test.msa-pap: 0.577
   Tatoeba-test.multi-multi: 0.586
   Tatoeba-test.pap-eng: 0.603
   Tatoeba-test.pap-msa: 0.197