opus-2020-06-28.yml

release: eng-inc/opus-2020-06-28.zip
release-date: 2020-06-28
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
   - sin
   - snd
   - urd
use-target-labels:
   - >>asm<<
   - >>awa<<
   - >>ben<<
   - >>bho<<
   - >>gom<<
   - >>guj<<
   - >>hif_Latn<<
   - >>hin<<
   - >>mai<<
   - >>mar<<
   - >>npi<<
   - >>ori<<
   - >>pan_Guru<<
   - >>pnb<<
   - >>rom<<
   - >>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: 0/0
   - 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-sin: 45/189
   - Tatoeba-test.eng-snd: 4/17
   - Tatoeba-test.eng-urd: 1663/12106
BLEU-scores:
   - Tatoeba-test.eng-asm: 3.0
   - Tatoeba-test.eng-awa: 0.4
   - Tatoeba-test.eng-ben: 16.5
   - Tatoeba-test.eng-bho: 0.8
   - Tatoeba-test.eng-guj: 19.9
   - Tatoeba-test.eng-hif: 0.5
   - Tatoeba-test.eng-hin: 17.4
   - Tatoeba-test.eng-kok: 8.1
   - Tatoeba-test.eng-lah: 0.2
   - Tatoeba-test.eng-mai: 7.6
   - Tatoeba-test.eng-mar: 20.4
   - Tatoeba-test.eng-multi: 17.0
   - Tatoeba-test.eng-nep: 1.0
   - Tatoeba-test.eng-ori: 2.2
   - Tatoeba-test.eng-pan: 8.4
   - Tatoeba-test.eng-rom: 0.3
   - Tatoeba-test.eng-sin: 9.5
   - Tatoeba-test.eng-snd: 6.8
   - Tatoeba-test.eng-urd: 12.5
chr-F-scores:
   - Tatoeba-test.eng-asm: 0.245
   - Tatoeba-test.eng-awa: 0.098
   - Tatoeba-test.eng-ben: 0.481
   - Tatoeba-test.eng-bho: 0.110
   - Tatoeba-test.eng-guj: 0.393
   - Tatoeba-test.eng-hif: 0.022
   - Tatoeba-test.eng-hin: 0.463
   - Tatoeba-test.eng-kok: 0.006
   - Tatoeba-test.eng-lah: 0.001
   - Tatoeba-test.eng-mai: 0.374
   - Tatoeba-test.eng-mar: 0.464
   - Tatoeba-test.eng-multi: 0.442
   - Tatoeba-test.eng-nep: 0.102
   - Tatoeba-test.eng-ori: 0.198
   - Tatoeba-test.eng-pan: 0.343
   - Tatoeba-test.eng-rom: 0.185
   - Tatoeba-test.eng-sin: 0.368
   - Tatoeba-test.eng-snd: 0.343
   - Tatoeba-test.eng-urd: 0.414