英语(en)翻译成西班牙语(es)的神经机器翻译模型。
该模型是 OPUS-MT project 项目的一部分,旨在让神经机器翻译模型广泛可用和易于访问世界上的许多语言。所有模型最初使用纯 C++ 编写的 Marian NMT 框架进行训练。使用 transformers 库由 huggingface 转换为 pyTorch。训练数据取自 OPUS ,并且训练流程使用 OPUS-MT-train 的程序。
@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",
}
简短的示例代码:
from transformers import MarianMTModel, MarianTokenizer
src_text = [
"A wasp stung him and he had an allergic reaction.",
"I love nature."
]
model_name = "pytorch-models/opus-mt-tc-big-en-es"
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:
# Una avispa lo picó y tuvo una reacción alérgica.
# Me encanta la naturaleza.
您还可以使用 transformers pipelines 使用 OPUS-MT 模型,例如:
from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-en-es")
print(pipe("A wasp stung him and he had an allergic reaction."))
# expected output: Una avispa lo picó y tuvo una reacción alérgica.
| langpair | testset | chr-F | BLEU | #sent | #words |
|---|---|---|---|---|---|
| eng-spa | tatoeba-test-v2021-08-07 | 0.73863 | 57.2 | 16583 | 134710 |
| eng-spa | flores101-devtest | 0.56440 | 28.5 | 1012 | 29199 |
| eng-spa | newssyscomb2009 | 0.58415 | 31.5 | 502 | 12503 |
| eng-spa | news-test2008 | 0.56707 | 30.1 | 2051 | 52586 |
| eng-spa | newstest2009 | 0.57836 | 30.2 | 2525 | 68111 |
| eng-spa | newstest2010 | 0.62357 | 37.6 | 2489 | 65480 |
| eng-spa | newstest2011 | 0.62415 | 38.9 | 3003 | 79476 |
| eng-spa | newstest2012 | 0.63031 | 39.5 | 3003 | 79006 |
| eng-spa | newstest2013 | 0.60354 | 35.9 | 3000 | 70528 |
| eng-spa | tico19-test | 0.73554 | 53.0 | 2100 | 66563 |
这项工作得到 European Language Grid 支持,作为 pilot project 2866 的一部分,由 FoTran project 资助,该资助是欧洲研究理事会(ERC)在欧洲联盟的Horizon 2020研究和创新计划(授予 No 771113)下的一部分,以及 MeMAD project ,该资助是欧洲联盟Horizon 2020研究和创新计划(授予 No 780069)下的一部分。我们还感谢 CSC -- IT Center for Science 为芬兰提供的慷慨计算资源和IT基础设施。