opus-2020-10-04.yml

release: sit-sit/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:
   - bod
   - brx
   - cjy
   - cmn
   - eng
   - gan
   - hak
   - hsn
   - lzh
   - mya
   - nan
   - wuu
   - yue
target-languages:
   - bod
   - brx
   - cjy
   - cmn
   - eng
   - gan
   - hak
   - hsn
   - lzh
   - mya
   - nan
   - wuu
   - yue
use-target-labels:
   - >>bod<<
   - >>brx<<
   - >>brx_Latn<<
   - >>cjy_Hans<<
   - >>cjy_Hant<<
   - >>cmn<<
   - >>cmn_Hans<<
   - >>cmn_Hant<<
   - >>eng<<
   - >>gan<<
   - >>hak<<
   - >>hak_Hani<<
   - >>hsn_Hani<<
   - >>lzh<<
   - >>lzh_Hans<<
   - >>mya<<
   - >>nan<<
   - >>wuu<<
   - >>yue_Hans<<
   - >>yue_Hant<<
test-data:
   - Tatoeba-test.bod-eng: 38/186
   - Tatoeba-test.brx-eng: 12/46
   - Tatoeba-test.eng-bod: 38/42
   - Tatoeba-test.eng-brx: 12/40
   - Tatoeba-test.eng-mya: 216/1196
   - Tatoeba-test.eng-zho: 10000/10211
   - Tatoeba-test.multi-multi: 10000/41175
   - Tatoeba-test.mya-eng: 216/1427
   - Tatoeba-test.mya-zho: 17/18
   - Tatoeba-test.zho-eng: 10000/70665
   - Tatoeba-test.zho-mya: 17/68
   - Tatoeba-test.zho-zho: 2500/2515
BLEU-scores:
   - Tatoeba-test.bod-eng: 2.2
   - Tatoeba-test.brx-eng: 2.8
   - Tatoeba-test.eng-bod: 3.1
   - Tatoeba-test.eng-brx: 0.5
   - Tatoeba-test.eng-mya: 0.4
   - Tatoeba-test.eng-zho: 28.1
   - Tatoeba-test.multi-multi: 28.4
   - Tatoeba-test.mya-eng: 15.6
   - Tatoeba-test.mya-zho: 3.6
   - Tatoeba-test.zho-eng: 29.8
   - Tatoeba-test.zho-mya: 0.4
   - Tatoeba-test.zho-zho: 13.1
chr-F-scores:
   - Tatoeba-test.bod-eng: 0.135
   - Tatoeba-test.brx-eng: 0.134
   - Tatoeba-test.eng-bod: 0.132
   - Tatoeba-test.eng-brx: 0.071
   - Tatoeba-test.eng-mya: 0.160
   - Tatoeba-test.eng-zho: 0.239
   - Tatoeba-test.multi-multi: 0.428
   - Tatoeba-test.mya-eng: 0.361
   - Tatoeba-test.mya-zho: 0.056
   - Tatoeba-test.zho-eng: 0.495
   - Tatoeba-test.zho-mya: 0.146
   - Tatoeba-test.zho-zho: 0.099