README.md
3.2 KB · 110 lines · markdown Raw
1 ---
2 tags:
3 - translation
4 license: cc-by-4.0
5 ---
6
7 ### opus-mt-en-de
8
9
10 ## Table of Contents
11 - [Model Details](#model-details)
12 - [Uses](#uses)
13 - [Risks, Limitations and Biases](#risks-limitations-and-biases)
14 - [Training](#training)
15 - [Evaluation](#evaluation)
16 - [Citation Information](#citation-information)
17 - [How to Get Started With the Model](#how-to-get-started-with-the-model)
18
19 ## Model Details
20 **Model Description:**
21 - **Developed by:** Language Technology Research Group at the University of Helsinki
22 - **Model Type:** Translation
23 - **Language(s):**
24 - Source Language: English
25 - Target Language: German
26 - **License:** CC-BY-4.0
27 - **Resources for more information:**
28 - [GitHub Repo](https://github.com/Helsinki-NLP/OPUS-MT-train)
29
30
31 ## Uses
32
33 #### Direct Use
34
35 This model can be used for translation and text-to-text generation.
36
37
38 ## Risks, Limitations and Biases
39
40
41
42 **CONTENT WARNING: Readers should be aware this section contains content that is disturbing, offensive, and can propagate historical and current stereotypes.**
43
44 Significant research has explored bias and fairness issues with language models (see, e.g., [Sheng et al. (2021)](https://aclanthology.org/2021.acl-long.330.pdf) and [Bender et al. (2021)](https://dl.acm.org/doi/pdf/10.1145/3442188.3445922)).
45
46 Further details about the dataset for this model can be found in the OPUS readme: [en-de](https://github.com/Helsinki-NLP/OPUS-MT-train/blob/master/models/en-de/README.md)
47
48
49 #### Training Data
50 ##### Preprocessing
51 * pre-processing: normalization + SentencePiece
52
53 * dataset: [opus](https://github.com/Helsinki-NLP/Opus-MT)
54 * download original weights: [opus-2020-02-26.zip](https://object.pouta.csc.fi/OPUS-MT-models/en-de/opus-2020-02-26.zip)
55
56 * test set translations: [opus-2020-02-26.test.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-de/opus-2020-02-26.test.txt)
57
58 ## Evaluation
59
60 #### Results
61
62 * test set scores: [opus-2020-02-26.eval.txt](https://object.pouta.csc.fi/OPUS-MT-models/en-de/opus-2020-02-26.eval.txt)
63
64
65 #### Benchmarks
66
67 | testset | BLEU | chr-F |
68 |-----------------------|-------|-------|
69 | newssyscomb2009.en.de | 23.5 | 0.540 |
70 | news-test2008.en.de | 23.5 | 0.529 |
71 | newstest2009.en.de | 22.3 | 0.530 |
72 | newstest2010.en.de | 24.9 | 0.544 |
73 | newstest2011.en.de | 22.5 | 0.524 |
74 | newstest2012.en.de | 23.0 | 0.525 |
75 | newstest2013.en.de | 26.9 | 0.553 |
76 | newstest2015-ende.en.de | 31.1 | 0.594 |
77 | newstest2016-ende.en.de | 37.0 | 0.636 |
78 | newstest2017-ende.en.de | 29.9 | 0.586 |
79 | newstest2018-ende.en.de | 45.2 | 0.690 |
80 | newstest2019-ende.en.de | 40.9 | 0.654 |
81 | Tatoeba.en.de | 47.3 | 0.664 |
82
83
84
85 ## Citation Information
86
87 ```bibtex
88 @InProceedings{TiedemannThottingal:EAMT2020,
89 author = {J{\"o}rg Tiedemann and Santhosh Thottingal},
90 title = {{OPUS-MT} — {B}uilding open translation services for the {W}orld},
91 booktitle = {Proceedings of the 22nd Annual Conferenec of the European Association for Machine Translation (EAMT)},
92 year = {2020},
93 address = {Lisbon, Portugal}
94 }
95 ```
96
97 ## How to Get Started With the Model
98 ```python
99 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
100
101 tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-de")
102
103 model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-de")
104
105 ```
106
107
108
109
110