mirror of
https://github.com/Helsinki-NLP/OPUS-MT-train.git
synced 2024-10-26 21:19:02 +03:00
.. | ||
README.md |
opus-2019-12-04.zip
- dataset: opus
- model: transformer
- pre-processing: normalization + tokenization + BPE
- download: opus-2019-12-04.zip
- test set translations: opus-2019-12-04.test.txt
- test set scores: opus-2019-12-04.eval.txt
Benchmarks
testset | BLEU | chr-F |
---|---|---|
newssyscomb2009.de.en | 26.9 | 0.535 |
news-test2008.de.en | 24.8 | 0.519 |
newstest2009.de.en | 24.6 | 0.519 |
newstest2010.de.en | 27.5 | 0.552 |
newstest2011.de.en | 25.6 | 0.526 |
newstest2012.de.en | 27.1 | 0.535 |
newstest2013.de.en | 29.4 | 0.551 |
newstest2014-deen.de.en | 29.1 | 0.553 |
newstest2015-ende.de.en | 29.7 | 0.556 |
newstest2016-ende.de.en | 34.8 | 0.600 |
newstest2017-ende.de.en | 29.9 | 0.563 |
newstest2018-ende.de.en | 37.4 | 0.611 |
newstest2019-deen.de.en | 34.3 | 0.587 |
Tatoeba.de.en | 54.5 | 0.677 |
opus-2019-12-18.zip
- dataset: opus
- model: transformer-align
- pre-processing: normalization + SentencePiece
- download: opus-2019-12-18.zip
- test set translations: opus-2019-12-18.test.txt
- test set scores: opus-2019-12-18.eval.txt
Benchmarks
testset | BLEU | chr-F |
---|---|---|
newssyscomb2009.de.en | 28.6 | 0.553 |
news-test2008.de.en | 27.6 | 0.547 |
newstest2009.de.en | 26.9 | 0.544 |
newstest2010.de.en | 30.4 | 0.585 |
newstest2011.de.en | 27.5 | 0.554 |
newstest2012.de.en | 29.0 | 0.567 |
newstest2013.de.en | 32.2 | 0.583 |
newstest2014-deen.de.en | 33.8 | 0.596 |
newstest2015-ende.de.en | 34.3 | 0.598 |
newstest2016-ende.de.en | 40.1 | 0.646 |
newstest2017-ende.de.en | 35.6 | 0.609 |
newstest2018-ende.de.en | 43.8 | 0.667 |
newstest2019-deen.de.en | 39.6 | 0.637 |
Tatoeba.de.en | 55.1 | 0.704 |
opus-2020-02-26.zip
- dataset: opus
- model: transformer-align
- pre-processing: normalization + SentencePiece
- download: opus-2020-02-26.zip
- test set translations: opus-2020-02-26.test.txt
- test set scores: opus-2020-02-26.eval.txt
Benchmarks
testset | BLEU | chr-F |
---|---|---|
newssyscomb2009.de.en | 29.4 | 0.557 |
news-test2008.de.en | 27.8 | 0.548 |
newstest2009.de.en | 26.8 | 0.543 |
newstest2010.de.en | 30.2 | 0.584 |
newstest2011.de.en | 27.4 | 0.556 |
newstest2012.de.en | 29.1 | 0.569 |
newstest2013.de.en | 32.1 | 0.583 |
newstest2014-deen.de.en | 34.0 | 0.600 |
newstest2015-ende.de.en | 34.2 | 0.599 |
newstest2016-ende.de.en | 40.4 | 0.649 |
newstest2017-ende.de.en | 35.7 | 0.610 |
newstest2018-ende.de.en | 43.7 | 0.667 |
newstest2019-deen.de.en | 40.1 | 0.642 |
Tatoeba.de.en | 55.4 | 0.707 |