opus-mt-tc-big-zlw-en

Neural machine translation model for translating from West Slavic languages (zlw) 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.

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: 2022-03-17
source language(s): ces dsb hsb pol
target language(s): eng
model: transformer-big
data: opusTCv20210807+bt (source)
tokenization: SentencePiece (spm32k,spm32k)
original model: opusTCv20210807+bt_transformer-big_2022-03-17.zip
more information released models: OPUS-MT zlw-eng README

Usage

A short example code:

from transformers import MarianMTModel, MarianTokenizer

src_text = [
    "Aoi'ego hobby to tańczenie.",
    "Myślisz, że Tom planuje to zrobić?"
]

model_name = "pytorch-models/opus-mt-tc-big-zlw-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:
#     Aoi's hobby is dancing.
#     You think Tom's planning on doing that?

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-zlw-en")
print(pipe("Aoi'ego hobby to tańczenie."))

# expected output: Aoi's hobby is dancing.

Benchmarks

test set translations: opusTCv20210807+bt_transformer-big_2022-03-17.test.txt
test set scores: opusTCv20210807+bt_transformer-big_2022-03-17.eval.txt
benchmark results: benchmark_results.txt
benchmark output: benchmark_translations.zip

langpair	testset	chr-F	BLEU	#sent	#words
ces-eng	tatoeba-test-v2021-08-07	0.71861	57.4	13824	105010
pol-eng	tatoeba-test-v2021-08-07	0.70544	55.7	10099	75766
ces-eng	flores101-devtest	0.66444	41.2	1012	24721
pol-eng	flores101-devtest	0.58301	29.6	1012	24721
slk-eng	flores101-devtest	0.66103	40.0	1012	24721
ces-eng	multi30k_test_2016_flickr	0.61482	37.6	1000	12955
ces-eng	multi30k_test_2018_flickr	0.61405	37.4	1071	14689
pol-eng	newsdev2020	0.60478	32.7	2000	46654
ces-eng	newssyscomb2009	0.56495	30.2	502	11818
ces-eng	news-test2008	0.54300	26.3	2051	49380
ces-eng	newstest2009	0.56309	29.5	2525	65399
ces-eng	newstest2010	0.57778	30.7	2489	61711
ces-eng	newstest2011	0.57336	30.9	3003	74681
ces-eng	newstest2012	0.56761	29.4	3003	72812
ces-eng	newstest2013	0.58809	32.8	3000	64505
ces-eng	newstest2014	0.64401	38.7	3003	68065
ces-eng	newstest2015	0.58607	33.4	2656	53569
ces-eng	newstest2016	0.61780	37.1	2999	64670
ces-eng	newstest2017	0.58259	32.5	3005	61721
ces-eng	newstest2018	0.58677	33.1	2983	63495
pol-eng	newstest2020	0.60047	32.6	1001	21755

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: 3405783
port time: Wed Apr 13 20:19:48 EEST 2022
port machine: LM0-400-22516.local

Helsinki-NLP
/

opus-mt-tc-big-zlw-en