itc-deu

opus1m-2021-05-16.zip

• dataset: opus1m
• model: transformer-align
• source language(s): ast cat cos mol ron
• target language(s): deu
• model: transformer-align
• pre-processing: normalization + SentencePiece (spm32k,spm32k)
• download: opus1m-2021-05-16.zip
• test set translations: opus1m-2021-05-16.test.txt
• test set scores: opus1m-2021-05-16.eval.txt

Benchmarks

testset	BLEU	chr-F	#sent	#words	BP
Tatoeba-test.ast-deu	16.7	0.340	34	253	0.964
Tatoeba-test.cat-deu	39.6	0.595	720	5640	0.990
Tatoeba-test.cos-deu	0.0	0.048	1	2	1.000
Tatoeba-test.egl-deu	3.6	0.095	3	15	1.000
Tatoeba-test.fra-deu	2.6	0.194	10000	80251	1.000
Tatoeba-test.glg-deu	4.6	0.255	103	688	1.000
Tatoeba-test.ita-deu	6.3	0.252	10000	78689	0.990
Tatoeba-test.lad-deu	8.7	0.239	173	910	1.000
Tatoeba-test.lat-deu	2.8	0.175	2016	13323	0.855
Tatoeba-test.multi-deu	41.0	0.596	1896	13791	0.986
Tatoeba-test.oci-deu	6.1	0.255	174	1319	1.000
Tatoeba-test.osp-deu	23.2	0.370	3	15	1.000
Tatoeba-test.pms-deu	1.0	0.145	12	91	1.000
Tatoeba-test.por-deu	4.4	0.235	10000	81221	1.000
Tatoeba-test.roh-deu	8.1	0.234	14	141	0.896
Tatoeba-test.ron-deu	40.8	0.590	1141	7889	0.975
Tatoeba-test.scn-deu	2.3	0.140	4	31	0.825
Tatoeba-test.spa-deu	13.6	0.348	10000	81191	1.000