opus-2020-10-04.yml

release: dra-dra/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
   - kan
   - mal
   - tam
   - tel
target-languages:
   - eng
   - kan
   - mal
   - tam
   - tel
use-target-labels:
   - ">>eng<<"
   - ">>kan<<"
   - ">>mal<<"
   - ">>tam<<"
   - ">>tam_Deva<<"
   - ">>tel<<"
test-data:
   Tatoeba-test.eng-kan: 167/783
   Tatoeba-test.eng-mal: 802/3459
   Tatoeba-test.eng-tam: 311/1364
   Tatoeba-test.eng-tel: 261/1085
   Tatoeba-test.kan-eng: 167/1073
   Tatoeba-test.mal-eng: 802/4700
   Tatoeba-test.multi-multi: 3082/15652
   Tatoeba-test.tam-eng: 311/1787
   Tatoeba-test.tel-eng: 261/1401
BLEU-scores:
   Tatoeba-test.eng-kan: 7.4
   Tatoeba-test.eng-mal: 13.2
   Tatoeba-test.eng-tam: 9.2
   Tatoeba-test.eng-tel: 7.1
   Tatoeba-test.kan-eng: 22.6
   Tatoeba-test.mal-eng: 43.2
   Tatoeba-test.multi-multi: 26.8
   Tatoeba-test.tam-eng: 31.1
   Tatoeba-test.tel-eng: 24.2
chr-F-scores:
   Tatoeba-test.eng-kan: 0.376
   Tatoeba-test.eng-mal: 0.499
   Tatoeba-test.eng-tam: 0.434
   Tatoeba-test.eng-tel: 0.364
   Tatoeba-test.kan-eng: 0.435
   Tatoeba-test.mal-eng: 0.600
   Tatoeba-test.multi-multi: 0.486
   Tatoeba-test.tam-eng: 0.466
   Tatoeba-test.tel-eng: 0.430