README.md
3.1 KB · 108 lines · markdown Raw
1 ---
2 language:
3 - zh
4 - en
5
6 tags:
7 - translation
8
9 license: cc-by-4.0
10 ---
11
12 ### zho-eng
13
14 ## Table of Contents
15 - [Model Details](#model-details)
16 - [Uses](#uses)
17 - [Risks, Limitations and Biases](#risks-limitations-and-biases)
18 - [Training](#training)
19 - [Evaluation](#evaluation)
20 - [Citation Information](#citation-information)
21 - [How to Get Started With the Model](#how-to-get-started-with-the-model)
22
23 ## Model Details
24 - **Model Description:**
25 - **Developed by:** Language Technology Research Group at the University of Helsinki
26 - **Model Type:** Translation
27 - **Language(s):**
28 - Source Language: Chinese
29 - Target Language: English
30 - **License:** CC-BY-4.0
31 - **Resources for more information:**
32 - [GitHub Repo](https://github.com/Helsinki-NLP/OPUS-MT-train)
33
34
35 ## Uses
36
37 #### Direct Use
38
39 This model can be used for translation and text-to-text generation.
40
41
42 ## Risks, Limitations and Biases
43
44 **CONTENT WARNING: Readers should be aware this section contains content that is disturbing, offensive, and can propagate historical and current stereotypes.**
45
46 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)).
47
48 Further details about the dataset for this model can be found in the OPUS readme: [zho-eng](https://github.com/Helsinki-NLP/Tatoeba-Challenge/tree/master/models/zho-eng/README.md)
49
50 ## Training
51
52 #### System Information
53 * helsinki_git_sha: 480fcbe0ee1bf4774bcbe6226ad9f58e63f6c535
54 * transformers_git_sha: 2207e5d8cb224e954a7cba69fa4ac2309e9ff30b
55 * port_machine: brutasse
56 * port_time: 2020-08-21-14:41
57 * src_multilingual: False
58 * tgt_multilingual: False
59
60 #### Training Data
61 ##### Preprocessing
62 * pre-processing: normalization + SentencePiece (spm32k,spm32k)
63 * ref_len: 82826.0
64 * dataset: [opus](https://github.com/Helsinki-NLP/Opus-MT)
65 * download original weights: [opus-2020-07-17.zip](https://object.pouta.csc.fi/Tatoeba-MT-models/zho-eng/opus-2020-07-17.zip)
66
67 * test set translations: [opus-2020-07-17.test.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/zho-eng/opus-2020-07-17.test.txt)
68
69
70 ## Evaluation
71
72 #### Results
73
74 * test set scores: [opus-2020-07-17.eval.txt](https://object.pouta.csc.fi/Tatoeba-MT-models/zho-eng/opus-2020-07-17.eval.txt)
75
76 * brevity_penalty: 0.948
77
78
79 ## Benchmarks
80
81 | testset | BLEU | chr-F |
82 |-----------------------|-------|-------|
83 | Tatoeba-test.zho.eng | 36.1 | 0.548 |
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
99 ```python
100 from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
101
102 tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-zh-en")
103
104 model = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-zh-en")
105 ```
106
107
108