opus-2020-07-06.yml

release: afa-afa/opus-2020-07-06.zip
release-date: 2020-07-06
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:
   - apc
   - ara
   - arq
   - arz
   - heb
   - kab
   - mlt
   - shy
   - thv
target-languages:
   - apc
   - ara
   - arq
   - arz
   - heb
   - kab
   - mlt
   - shy
   - thv
use-target-labels:
   - ">>apc<<"
   - ">>ara<<"
   - ">>arq<<"
   - ">>arz<<"
   - ">>heb<<"
   - ">>kab<<"
   - ">>mlt<<"
   - ">>shy_Latn<<"
   - ">>thv<<"
test-data:
   Tatoeba-test.ara-ara: 16/48
   Tatoeba-test.ara-heb: 1208/5440
   Tatoeba-test.ara-kab: 147/643
   Tatoeba-test.ara-mlt: 28/59
   Tatoeba-test.ara-shy: 15/49
   Tatoeba-test.heb-ara: 1208/5301
   Tatoeba-test.heb-kab: 3/8
   Tatoeba-test.kab-ara: 147/591
   Tatoeba-test.kab-heb: 3/7
   Tatoeba-test.kab-shy: 3/23
   Tatoeba-test.kab-tmh: 1/3
   Tatoeba-test.mlt-ara: 28/67
   Tatoeba-test.multi-multi: 2842/12365
   Tatoeba-test.shy-ara: 15/52
   Tatoeba-test.shy-kab: 3/23
   Tatoeba-test.tmh-kab: 1/3
BLEU-scores:
   Tatoeba-test.ara-ara: 0.9
   Tatoeba-test.ara-heb: 30.7
   Tatoeba-test.ara-kab: 0.2
   Tatoeba-test.ara-mlt: 13.6
   Tatoeba-test.ara-shy: 1.1
   Tatoeba-test.heb-ara: 14.4
   Tatoeba-test.heb-kab: 2.2
   Tatoeba-test.kab-ara: 0.3
   Tatoeba-test.kab-heb: 7.0
   Tatoeba-test.kab-shy: 0.8
   Tatoeba-test.kab-tmh: 4.1
   Tatoeba-test.mlt-ara: 27.7
   Tatoeba-test.multi-multi: 19.6
   Tatoeba-test.shy-ara: 1.3
   Tatoeba-test.shy-kab: 1.3
   Tatoeba-test.tmh-kab: 2.8
chr-F-scores:
   Tatoeba-test.ara-ara: 0.064
   Tatoeba-test.ara-heb: 0.497
   Tatoeba-test.ara-kab: 0.108
   Tatoeba-test.ara-mlt: 0.429
   Tatoeba-test.ara-shy: 0.042
   Tatoeba-test.heb-ara: 0.422
   Tatoeba-test.heb-kab: 0.127
   Tatoeba-test.kab-ara: 0.094
   Tatoeba-test.kab-heb: 0.083
   Tatoeba-test.kab-shy: 0.000
   Tatoeba-test.kab-tmh: 0.000
   Tatoeba-test.mlt-ara: 0.425
   Tatoeba-test.multi-multi: 0.404
   Tatoeba-test.shy-ara: 0.080
   Tatoeba-test.shy-kab: 0.013
   Tatoeba-test.tmh-kab: 0.077