opus-2020-07-04.yml

release: gem-eng/opus-2020-07-04.zip
release-date: 2020-07-04
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:
   - afr
   - ang
   - dan
   - deu
   - enm
   - fao
   - frr
   - fry
   - gos
   - got
   - gsw
   - isl
   - ksh
   - ltz
   - nds
   - nld
   - non
   - pdc
   - sco
   - stq
   - swe
   - swg
   - yid
target-languages:
   - eng
test-data:
   - Tatoeba-test.afr-eng: 1374/8101
   - Tatoeba-test.ang-eng: 189/1693
   - Tatoeba-test.dan-eng: 10000/62015
   - Tatoeba-test.deu-eng: 10000/69484
   - Tatoeba-test.enm-eng: 49/238
   - Tatoeba-test.fao-eng: 294/1631
   - Tatoeba-test.frr-eng: 2/7
   - Tatoeba-test.fry-eng: 205/1234
   - Tatoeba-test.gos-eng: 1152/4319
   - Tatoeba-test.got-eng: 207/865
   - Tatoeba-test.gsw-eng: 205/770
   - Tatoeba-test.isl-eng: 2500/16860
   - Tatoeba-test.ksh-eng: 26/175
   - Tatoeba-test.ltz-eng: 283/1378
   - Tatoeba-test.multi-eng: 10000/60720
   - Tatoeba-test.nds-eng: 2500/14803
   - Tatoeba-test.nld-eng: 10000/58489
   - Tatoeba-test.non-eng: 15/112
   - Tatoeba-test.pdc-eng: 53/295
   - Tatoeba-test.sco-eng: 27/183
   - Tatoeba-test.stq-eng: 5/28
   - Tatoeba-test.swe-eng: 10000/54997
   - Tatoeba-test.swg-eng: 33/205
   - Tatoeba-test.yid-eng: 1168/6380
BLEU-scores:
   - Tatoeba-test.afr-eng: 61.5
   - Tatoeba-test.ang-eng: 7.0
   - Tatoeba-test.dan-eng: 60.3
   - Tatoeba-test.deu-eng: 48.7
   - Tatoeba-test.enm-eng: 23.0
   - Tatoeba-test.fao-eng: 21.6
   - Tatoeba-test.frr-eng: 14.1
   - Tatoeba-test.fry-eng: 27.5
   - Tatoeba-test.gos-eng: 17.0
   - Tatoeba-test.got-eng: 0.0
   - Tatoeba-test.gsw-eng: 13.4
   - Tatoeba-test.isl-eng: 50.2
   - Tatoeba-test.ksh-eng: 5.3
   - Tatoeba-test.ltz-eng: 34.1
   - Tatoeba-test.multi-eng: 53.0
   - Tatoeba-test.nds-eng: 31.5
   - Tatoeba-test.nld-eng: 58.0
   - Tatoeba-test.non-eng: 36.6
   - Tatoeba-test.pdc-eng: 26.7
   - Tatoeba-test.sco-eng: 44.9
   - Tatoeba-test.stq-eng: 5.0
   - Tatoeba-test.swe-eng: 61.5
   - Tatoeba-test.swg-eng: 20.7
   - Tatoeba-test.yid-eng: 17.2
chr-F-scores:
   - Tatoeba-test.afr-eng: 0.749
   - Tatoeba-test.ang-eng: 0.241
   - Tatoeba-test.dan-eng: 0.744
   - Tatoeba-test.deu-eng: 0.659
   - Tatoeba-test.enm-eng: 0.461
   - Tatoeba-test.fao-eng: 0.411
   - Tatoeba-test.frr-eng: 0.167
   - Tatoeba-test.fry-eng: 0.474
   - Tatoeba-test.gos-eng: 0.336
   - Tatoeba-test.got-eng: 0.004
   - Tatoeba-test.gsw-eng: 0.327
   - Tatoeba-test.isl-eng: 0.663
   - Tatoeba-test.ksh-eng: 0.242
   - Tatoeba-test.ltz-eng: 0.497
   - Tatoeba-test.multi-eng: 0.683
   - Tatoeba-test.nds-eng: 0.515
   - Tatoeba-test.nld-eng: 0.729
   - Tatoeba-test.non-eng: 0.550
   - Tatoeba-test.pdc-eng: 0.424
   - Tatoeba-test.sco-eng: 0.589
   - Tatoeba-test.stq-eng: 0.363
   - Tatoeba-test.swe-eng: 0.742
   - Tatoeba-test.swg-eng: 0.334
   - Tatoeba-test.yid-eng: 0.380