translation opus-mt-tc

opus-mt-tc-big-zls-en

Neural machine translation model for translating from South Slavic languages (zls) to English (en).

This model is part of the OPUS-MT project, an effort to make neural machine translation models widely available and accessible for many languages in the world. All models are originally trained using the amazing framework of Marian NMT, an efficient NMT implementation written in pure C++. The models have been converted to pyTorch using the transformers library by huggingface. Training data is taken from OPUS and training pipelines use the procedures of OPUS-MT-train.

@inproceedings{tiedemann-thottingal-2020-opus,
    title = "{OPUS}-{MT} {--} Building open translation services for the World",
    author = {Tiedemann, J{\"o}rg  and Thottingal, Santhosh},
    booktitle = "Proceedings of the 22nd Annual Conference of the European Association for Machine Translation",
    month = nov,
    year = "2020",
    address = "Lisboa, Portugal",
    publisher = "European Association for Machine Translation",
    url = "https://aclanthology.org/2020.eamt-1.61",
    pages = "479--480",
}

@inproceedings{tiedemann-2020-tatoeba,
    title = "The Tatoeba Translation Challenge {--} Realistic Data Sets for Low Resource and Multilingual {MT}",
    author = {Tiedemann, J{\"o}rg},
    booktitle = "Proceedings of the Fifth Conference on Machine Translation",
    month = nov,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.wmt-1.139",
    pages = "1174--1182",
}

Model info

Usage

A short example code:

from transformers import MarianMTModel, MarianTokenizer

src_text = [
    "Да не би случайно Том да остави Мери да кара колата?",
    "Какво е времето днес?"
]

model_name = "pytorch-models/opus-mt-tc-big-zls-en"
tokenizer = MarianTokenizer.from_pretrained(model_name)
model = MarianMTModel.from_pretrained(model_name)
translated = model.generate(**tokenizer(src_text, return_tensors="pt", padding=True))

for t in translated:
    print( tokenizer.decode(t, skip_special_tokens=True) )

# expected output:
#     Did Tom just let Mary drive the car?
#     What's the weather like today?

You can also use OPUS-MT models with the transformers pipelines, for example:

from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-zls-en")
print(pipe("Да не би случайно Том да остави Мери да кара колата?"))

# expected output: Did Tom just let Mary drive the car?

Benchmarks

langpair testset chr-F BLEU #sent #words
bos_Latn-eng tatoeba-test-v2021-08-07 0.79339 66.5 301 1826
bul-eng tatoeba-test-v2021-08-07 0.72656 59.3 10000 71872
hbs-eng tatoeba-test-v2021-08-07 0.71783 57.3 10017 68934
hrv-eng tatoeba-test-v2021-08-07 0.74066 59.2 1480 10620
mkd-eng tatoeba-test-v2021-08-07 0.70043 57.4 10010 65667
slv-eng tatoeba-test-v2021-08-07 0.39534 23.5 2495 16940
srp_Cyrl-eng tatoeba-test-v2021-08-07 0.67628 47.0 1580 10181
srp_Latn-eng tatoeba-test-v2021-08-07 0.71878 58.5 6656 46307
bul-eng flores101-devtest 0.67375 42.0 1012 24721
hrv-eng flores101-devtest 0.63914 37.1 1012 24721
mkd-eng flores101-devtest 0.67444 43.2 1012 24721
slv-eng flores101-devtest 0.62087 35.2 1012 24721
srp_Cyrl-eng flores101-devtest 0.67810 36.8 1012 24721

Acknowledgements

The work is supported by the European Language Grid as pilot project 2866, by the FoTran project, funded by the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement No 771113), and the MeMAD project, funded by the European Union’s Horizon 2020 Research and Innovation Programme under grant agreement No 780069. We are also grateful for the generous computational resources and IT infrastructure provided by CSC -- IT Center for Science, Finland.

Model conversion info