File size: 2,712 Bytes
2ff9a04 6c5075e 50a8304 6c5075e 50a8304 6c5075e 50a8304 6c5075e 50a8304 2ff9a04 6c5075e 50a8304 6c5075e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 |
---
language:
- en
- de
- multilingual
license: cc-by-4.0
tags:
- translation
- opus-mt-tc
model-index:
- name: opus-mt-tc-base-en-de
results:
- task:
type: translation
name: Translation eng-deu
dataset:
name: tatoeba-test-v2021-08-07
type: tatoeba_mt
args: eng-deu
metrics:
- type: bleu
value: 43.7
name: BLEU
---
# Opus Tatoeba English-German
*This model was obtained by running the script [convert_marian_to_pytorch.py](https://github.com/huggingface/transformers/blob/master/src/transformers/models/marian/convert_marian_to_pytorch.py). The original models were trained by [J�rg Tiedemann](https://blogs.helsinki.fi/tiedeman/) using the [MarianNMT](https://marian-nmt.github.io/) library. See all available `MarianMTModel` models on the profile of the [Helsinki NLP](https://huggingface.co/Helsinki-NLP) group.*
* dataset: opusTCv20210807+bt
* model: transformer-big
* source language(s): eng
* target language(s): deu
* raw source language(s): eng
* raw target language(s): deu
* model: transformer-big
* pre-processing: normalization + SentencePiece (spm32k,spm32k)
* download: [opusTCv20210807+bt-2021-12-08.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-deu/opusTCv20210807+bt-2021-12-08.zip)
* test set translations: [opusTCv20210807+bt-2021-12-08.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-deu/opusTCv20210807+bt-2021-12-08.test.txt)
* test set scores: [opusTCv20210807+bt-2021-12-08.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-deu/opusTCv20210807+bt-2021-12-08.eval.txt)
## Benchmarks
| testset | BLEU | chr-F | #sent | #words | BP |
|---------|-------|-------|-------|--------|----|
| newssyscomb2009.eng-deu | 24.3 | 0.5462 | 502 | 11271 | 0.993 |
| news-test2008.eng-deu | 24.7 | 0.5412 | 2051 | 47427 | 1.000 |
| newstest2009.eng-deu | 23.6 | 0.5385 | 2525 | 62816 | 0.999 |
| newstest2010.eng-deu | 26.9 | 0.5589 | 2489 | 61511 | 0.966 |
| newstest2011.eng-deu | 24.1 | 0.5364 | 3003 | 72981 | 0.990 |
| newstest2012.eng-deu | 24.6 | 0.5375 | 3003 | 72886 | 0.972 |
| newstest2013.eng-deu | 28.3 | 0.5636 | 3000 | 63737 | 0.988 |
| newstest2014-deen.eng-deu | 30.9 | 0.6084 | 3003 | 62964 | 1.000 |
| newstest2015-ende.eng-deu | 33.2 | 0.6106 | 2169 | 44260 | 1.000 |
| newstest2016-ende.eng-deu | 39.8 | 0.6595 | 2999 | 62670 | 0.993 |
| newstest2017-ende.eng-deu | 32.0 | 0.6047 | 3004 | 61291 | 1.000 |
| newstest2018-ende.eng-deu | 48.8 | 0.7146 | 2998 | 64276 | 1.000 |
| newstest2019-ende.eng-deu | 45.0 | 0.6821 | 1997 | 48969 | 0.995 |
| Tatoeba-test-v2021-08-07.eng-deu | 43.7 | 0.6442 | 10000 | 85728 | 1.000 |
|