opus-mt-en-urj / README.md
system
Update README.md 71c03b8
1
---
2
language: 
3
- en
4
- se
5
- fi
6
- hu
7
- et
8
- urj
9
10
tags:
11
- translation
12
13
license: apache-2.0
14
---
15
16
### eng-urj
17
18
* source group: English 
19
* target group: Uralic languages 
20
*  OPUS readme: [eng-urj](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-urj/README.md)
21
22
*  model: transformer
23
* source language(s): eng
24
* target language(s): est fin fkv_Latn hun izh kpv krl liv_Latn mdf mhr myv sma sme udm vro
25
* model: transformer
26
* pre-processing: normalization + SentencePiece (spm32k,spm32k)
27
* a sentence initial language token is required in the form of `>>id<<` (id = valid target language ID)
28
* download original weights: [opus2m-2020-08-02.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-urj/opus2m-2020-08-02.zip)
29
* test set translations: [opus2m-2020-08-02.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-urj/opus2m-2020-08-02.test.txt)
30
* test set scores: [opus2m-2020-08-02.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/eng-urj/opus2m-2020-08-02.eval.txt)
31
32
## Benchmarks
33
34
| testset               | BLEU  | chr-F |
35
|-----------------------|-------|-------|
36
| newsdev2015-enfi-engfin.eng.fin 	| 18.3 	| 0.519 |
37
| newsdev2018-enet-engest.eng.est 	| 19.3 	| 0.520 |
38
| newssyscomb2009-enghun.eng.hun 	| 15.4 	| 0.471 |
39
| newstest2009-enghun.eng.hun 	| 15.7 	| 0.468 |
40
| newstest2015-enfi-engfin.eng.fin 	| 20.2 	| 0.534 |
41
| newstest2016-enfi-engfin.eng.fin 	| 20.7 	| 0.541 |
42
| newstest2017-enfi-engfin.eng.fin 	| 23.6 	| 0.566 |
43
| newstest2018-enet-engest.eng.est 	| 20.8 	| 0.535 |
44
| newstest2018-enfi-engfin.eng.fin 	| 15.8 	| 0.499 |
45
| newstest2019-enfi-engfin.eng.fin 	| 19.9 	| 0.518 |
46
| newstestB2016-enfi-engfin.eng.fin 	| 16.6 	| 0.509 |
47
| newstestB2017-enfi-engfin.eng.fin 	| 19.4 	| 0.529 |
48
| Tatoeba-test.eng-chm.eng.chm 	| 1.3 	| 0.127 |
49
| Tatoeba-test.eng-est.eng.est 	| 51.0 	| 0.692 |
50
| Tatoeba-test.eng-fin.eng.fin 	| 34.6 	| 0.597 |
51
| Tatoeba-test.eng-fkv.eng.fkv 	| 2.2 	| 0.302 |
52
| Tatoeba-test.eng-hun.eng.hun 	| 35.6 	| 0.591 |
53
| Tatoeba-test.eng-izh.eng.izh 	| 5.7 	| 0.211 |
54
| Tatoeba-test.eng-kom.eng.kom 	| 3.0 	| 0.012 |
55
| Tatoeba-test.eng-krl.eng.krl 	| 8.5 	| 0.230 |
56
| Tatoeba-test.eng-liv.eng.liv 	| 2.7 	| 0.077 |
57
| Tatoeba-test.eng-mdf.eng.mdf 	| 2.8 	| 0.007 |
58
| Tatoeba-test.eng.multi 	| 35.1 	| 0.588 |
59
| Tatoeba-test.eng-myv.eng.myv 	| 1.3 	| 0.014 |
60
| Tatoeba-test.eng-sma.eng.sma 	| 1.8 	| 0.095 |
61
| Tatoeba-test.eng-sme.eng.sme 	| 6.8 	| 0.204 |
62
| Tatoeba-test.eng-udm.eng.udm 	| 1.1 	| 0.121 |
63
64
65
### System Info: 
66
- hf_name: eng-urj
67
68
- source_languages: eng
69
70
- target_languages: urj
71
72
- opus_readme_url: https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/eng-urj/README.md
73
74
- original_repo: Tatoeba-Challenge
75
76
- tags: ['translation']
77
78
- languages: ['en', 'se', 'fi', 'hu', 'et', 'urj']
79
80
- src_constituents: {'eng'}
81
82
- tgt_constituents: {'izh', 'mdf', 'vep', 'vro', 'sme', 'myv', 'fkv_Latn', 'krl', 'fin', 'hun', 'kpv', 'udm', 'liv_Latn', 'est', 'mhr', 'sma'}
83
84
- src_multilingual: False
85
86
- tgt_multilingual: True
87
88
- prepro:  normalization + SentencePiece (spm32k,spm32k)
89
90
- url_model: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-urj/opus2m-2020-08-02.zip
91
92
- url_test_set: https://object.pouta.csc.fi/Tatoeba-MT-models/eng-urj/opus2m-2020-08-02.test.txt
93
94
- src_alpha3: eng
95
96
- tgt_alpha3: urj
97
98
- short_pair: en-urj
99
100
- chrF2_score: 0.588
101
102
- bleu: 35.1
103
104
- brevity_penalty: 0.943
105
106
- ref_len: 59664.0
107
108
- src_name: English
109
110
- tgt_name: Uralic languages
111
112
- train_date: 2020-08-02
113
114
- src_alpha2: en
115
116
- tgt_alpha2: urj
117
118
- prefer_old: False
119
120
- long_pair: eng-urj
121
122
- helsinki_git_sha: 480fcbe0ee1bf4774bcbe6226ad9f58e63f6c535
123
124
- transformers_git_sha: 2207e5d8cb224e954a7cba69fa4ac2309e9ff30b
125
126
- port_machine: brutasse
127
128
- port_time: 2020-08-21-14:41