opus4m-2020-08-12.yml

release: sem-eng/opus4m-2020-08-12.zip
release-date: 2020-08-12
dataset-name: opus4m
modeltype: transformer
pre-processing: normalization + SentencePiece (spm32k,spm32k)
subwords:
   source: spm32k
   target: spm32k
subword-models:
   source: source.spm
   target: target.spm
source-languages:
   - acm
   - afb
   - amh
   - apc
   - ara
   - arq
   - ary
   - arz
   - heb
   - mlt
   - tir
target-languages:
   - eng
test-data:
   Tatoeba-test.amh-eng: 190/802
   Tatoeba-test.ara-eng: 10000/62449
   Tatoeba-test.heb-eng: 10000/62332
   Tatoeba-test.mlt-eng: 203/935
   Tatoeba-test.multi-eng: 10000/61612
   Tatoeba-test.tir-eng: 69/426
BLEU-scores:
   Tatoeba-test.amh-eng: 41.8
   Tatoeba-test.ara-eng: 39.4
   Tatoeba-test.heb-eng: 45.0
   Tatoeba-test.mlt-eng: 51.5
   Tatoeba-test.multi-eng: 42.3
   Tatoeba-test.tir-eng: 17.1
chr-F-scores:
   Tatoeba-test.amh-eng: 0.587
   Tatoeba-test.ara-eng: 0.570
   Tatoeba-test.heb-eng: 0.612
   Tatoeba-test.mlt-eng: 0.670
   Tatoeba-test.multi-eng: 0.592
   Tatoeba-test.tir-eng: 0.357