opus-mt-tc-big-zle-zle

Neural machine translation model for translating from East Slavic languages (zle) to East Slavic languages (zle).

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.

Publications: OPUS-MT – Building open translation services for the World and The Tatoeba Translation Challenge – Realistic Data Sets for Low Resource and Multilingual MT (Please, cite if you use this model.)

@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

Release: 2022-03-07
source language(s): bel rus ukr
target language(s): bel rus ukr
valid target language labels: >>bel<< >>rus<< >>ukr<<
model: transformer-big
data: opusTCv20210807+bt (source)
tokenization: SentencePiece (spm32k,spm32k)
original model: opusTCv20210807+bt_transformer-big_2022-03-07.zip
more information released models: OPUS-MT zle-zle README
more information about the model: MarianMT

This is a multilingual translation model with multiple target languages. A sentence initial language token is required in the form of >>id<< (id = valid target language ID), e.g. >>bel<<

Usage

A short example code:

from transformers import MarianMTModel, MarianTokenizer

src_text = [
    ">>ukr<< Кот мёртвый.",
    ">>bel<< Джон живе в Нью-Йорку."
]

model_name = "pytorch-models/opus-mt-tc-big-zle-zle"
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:
#     Кіт мертвий.
#     Джон жыве ў Нью-Йорку.

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-zle-zle")
print(pipe(">>ukr<< Кот мёртвый."))

# expected output: Кіт мертвий.

Benchmarks

test set translations: opusTCv20210807+bt_transformer-big_2022-03-07.test.txt
test set scores: opusTCv20210807+bt_transformer-big_2022-03-07.eval.txt
benchmark results: benchmark_results.txt
benchmark output: benchmark_translations.zip

langpair	testset	chr-F	BLEU	#sent	#words
bel-rus	tatoeba-test-v2021-08-07	0.82526	68.6	2500	18895
bel-ukr	tatoeba-test-v2021-08-07	0.81036	65.5	2355	15179
rus-bel	tatoeba-test-v2021-08-07	0.66943	50.3	2500	18756
rus-ukr	tatoeba-test-v2021-08-07	0.83639	70.1	10000	60212
ukr-bel	tatoeba-test-v2021-08-07	0.75368	58.9	2355	15175
ukr-rus	tatoeba-test-v2021-08-07	0.86806	75.7	10000	60387
bel-rus	flores101-devtest	0.47960	14.5	1012	23295
bel-ukr	flores101-devtest	0.47335	12.8	1012	22810
rus-ukr	flores101-devtest	0.55287	25.5	1012	22810
ukr-rus	flores101-devtest	0.56224	28.3	1012	23295

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

transformers version: 4.16.2
OPUS-MT git hash: 1bdabf7
port time: Thu Mar 24 00:15:39 EET 2022
port machine: LM0-400-22516.local

Helsinki-NLP
/

opus-mt-tc-big-zle-zle