opus-2021-05-16.yml

release: nld-zho/opus-2021-05-16.zip
release-date: 2021-05-16
dataset-name: opus
modeltype: transformer-align
vocabulary:
   source: opus.spm32k-spm32k.vocab.yml
   target: opus.spm32k-spm32k.vocab.yml
pre-processing: normalization + SentencePiece (spm32k,spm32k)
subwords:
   source: spm32k
   target: spm32k
subword-models:
   source: source.spm
   target: target.spm
source-languages:
   - nld
target-languages:
   - cmn
   - nan
   - yue
use-target-labels:
   - ">>cmn<<"
   - ">>cmn_Hans<<"
   - ">>cmn_Hant<<"
   - ">>nan<<"
   - ">>yue_Hans<<"
   - ">>yue_Hant<<"
training-data:
   nld-cmn: Tatoeba-train (212115) 
   nld-cmn_Hans: Tatoeba-train (5663839) 
   nld-cmn_Hant: Tatoeba-train (3914933) 
   nld-nan: Tatoeba-train (50) 
   nld-yue_Hans: Tatoeba-train (18) 
   nld-yue_Hant: Tatoeba-train (4459) 
validation-data:
   cmn-nld: Tatoeba-dev, 32
   cmn_Hans-nld: Tatoeba-dev, 559
   cmn_Hant-nld: Tatoeba-dev, 409
   total-size-shuffled: 1000
   devset-selected: top 1000  lines of Tatoeba-dev.src.shuffled
test-data:
   Tatoeba-test.nld-cmn_Hans: 134/1187
   Tatoeba-test.nld-cmn_Hant: 1513/16856
   Tatoeba-test.nld-cmn: 7/62
   Tatoeba-test.nld-zho: 1654/18108
BLEU-scores:
   Tatoeba-test.nld-cmn_Hans: 37.0
   Tatoeba-test.nld-cmn_Hant: 30.1
   Tatoeba-test.nld-cmn: 18.6
   Tatoeba-test.nld-zho: 30.5
chr-F-scores:
   Tatoeba-test.nld-cmn_Hans: 0.314
   Tatoeba-test.nld-cmn_Hant: 0.261
   Tatoeba-test.nld-cmn: 0.447
   Tatoeba-test.nld-zho: 0.265