opus-2020-07-26.yml

release: eng-inc/opus-2020-07-26.zip
release-date: 2020-07-26
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
target-languages:
   - asm
   - awa
   - ben
   - bho
   - gom
   - guj
   - hif
   - hin
   - mai
   - mar
   - npi
   - ori
   - pan
   - pnb
   - rom
   - san
   - sin
   - snd
   - urd
use-target-labels:
   - ">>asm<<"
   - ">>awa<<"
   - ">>ben<<"
   - ">>bho<<"
   - ">>gom<<"
   - ">>guj<<"
   - ">>hif_Latn<<"
   - ">>hin<<"
   - ">>mai<<"
   - ">>mar<<"
   - ">>npi<<"
   - ">>ori<<"
   - ">>pan_Guru<<"
   - ">>pnb<<"
   - ">>rom<<"
   - ">>san_Deva<<"
   - ">>sin<<"
   - ">>snd_Arab<<"
   - ">>urd<<"
test-data:
   Tatoeba-test.eng-asm: 117/530
   Tatoeba-test.eng-awa: 279/1087
   Tatoeba-test.eng-ben: 2500/10942
   Tatoeba-test.eng-bho: 42/224
   Tatoeba-test.eng-guj: 154/680
   Tatoeba-test.eng-hif: 36/177
   Tatoeba-test.eng-hin: 5000/31361
   Tatoeba-test.eng-kok: 1/5
   Tatoeba-test.eng-lah: 32/180
   Tatoeba-test.eng-mai: 8/16
   Tatoeba-test.eng-mar: 10000/47665
   Tatoeba-test.eng-multi: 10000/53001
   Tatoeba-test.eng-nep: 115/398
   Tatoeba-test.eng-ori: 33/190
   Tatoeba-test.eng-pan: 87/570
   Tatoeba-test.eng-rom: 671/4283
   Tatoeba-test.eng-san: 144/366
   Tatoeba-test.eng-sin: 45/189
   Tatoeba-test.eng-snd: 4/17
   Tatoeba-test.eng-urd: 1663/12106
   newsdev2014.eng-hin: 520/9310
   newsdev2019-engu.eng-guj: 1998/34842
   newstest2014-hien.eng-hin: 2507/57037
   newstest2019-engu.eng-guj: 998/19483
BLEU-scores:
   Tatoeba-test.eng-asm: 3.7
   Tatoeba-test.eng-awa: 0.4
   Tatoeba-test.eng-ben: 16.0
   Tatoeba-test.eng-bho: 0.6
   Tatoeba-test.eng-guj: 20.2
   Tatoeba-test.eng-hif: 0.5
   Tatoeba-test.eng-hin: 17.3
   Tatoeba-test.eng-kok: 3.3
   Tatoeba-test.eng-lah: 0.2
   Tatoeba-test.eng-mai: 9.3
   Tatoeba-test.eng-mar: 19.9
   Tatoeba-test.eng-multi: 16.6
   Tatoeba-test.eng-nep: 0.7
   Tatoeba-test.eng-ori: 2.2
   Tatoeba-test.eng-pan: 7.0
   Tatoeba-test.eng-rom: 0.4
   Tatoeba-test.eng-san: 1.7
   Tatoeba-test.eng-sin: 9.1
   Tatoeba-test.eng-snd: 5.6
   Tatoeba-test.eng-urd: 12.9
   newsdev2014.eng-hin: 7.5
   newsdev2019-engu.eng-guj: 6.3
   newstest2014-hien.eng-hin: 11.0
   newstest2019-engu.eng-guj: 7.1
chr-F-scores:
   Tatoeba-test.eng-asm: 0.260
   Tatoeba-test.eng-awa: 0.144
   Tatoeba-test.eng-ben: 0.466
   Tatoeba-test.eng-bho: 0.143
   Tatoeba-test.eng-guj: 0.375
   Tatoeba-test.eng-hif: 0.040
   Tatoeba-test.eng-hin: 0.462
   Tatoeba-test.eng-kok: 0.044
   Tatoeba-test.eng-lah: 0.005
   Tatoeba-test.eng-mai: 0.385
   Tatoeba-test.eng-mar: 0.461
   Tatoeba-test.eng-multi: 0.436
   Tatoeba-test.eng-nep: 0.067
   Tatoeba-test.eng-ori: 0.196
   Tatoeba-test.eng-pan: 0.342
   Tatoeba-test.eng-rom: 0.187
   Tatoeba-test.eng-san: 0.109
   Tatoeba-test.eng-sin: 0.365
   Tatoeba-test.eng-snd: 0.343
   Tatoeba-test.eng-urd: 0.411
   newsdev2014.eng-hin: 0.337
   newsdev2019-engu.eng-guj: 0.282
   newstest2014-hien.eng-hin: 0.358
   newstest2019-engu.eng-guj: 0.291