神经机器翻译模型,用于将英语(en)翻译成捷克语和斯洛伐克语(ces+slk)。
该模型是 OPUS-MT project 的一部分,旨在使神经机器翻译模型在世界上的许多语言中得以广泛可用和可访问。所有模型最初是使用 Marian NMT 的惊人框架进行训练的, Marian NMT 是用纯C++编写的高效NMT实现。这些模型已经转换为pyTorch,使用了huggingface的transformers库。训练数据来自 OPUS ,并且训练流程使用了 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 = [
">>ces<< We were enemies.",
">>ces<< Do you think Tom knows what's going on?"
]
model_name = "pytorch-models/opus-mt-tc-big-en-ces_slk"
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:
# Byli jsme nepřátelé.
# Myslíš, že Tom ví, co se děje?
您还可以使用transformers pipelines来使用OPUS-MT模型,例如:
from transformers import pipeline
pipe = pipeline("translation", model="Helsinki-NLP/opus-mt-tc-big-en-ces_slk")
print(pipe(">>ces<< We were enemies."))
# expected output: Byli jsme nepřátelé.
| langpair | testset | chr-F | BLEU | #sent | #words |
|---|---|---|---|---|---|
| eng-ces | tatoeba-test-v2021-08-07 | 0.66128 | 47.5 | 13824 | 91332 |
| eng-ces | flores101-devtest | 0.60411 | 34.1 | 1012 | 22101 |
| eng-slk | flores101-devtest | 0.62415 | 35.9 | 1012 | 22543 |
| eng-ces | multi30k_test_2016_flickr | 0.58547 | 33.4 | 1000 | 10503 |
| eng-ces | multi30k_test_2018_flickr | 0.59236 | 33.4 | 1071 | 11631 |
| eng-ces | newssyscomb2009 | 0.52702 | 25.3 | 502 | 10032 |
| eng-ces | news-test2008 | 0.50286 | 22.8 | 2051 | 42484 |
| eng-ces | newstest2009 | 0.52152 | 24.3 | 2525 | 55533 |
| eng-ces | newstest2010 | 0.52527 | 24.4 | 2489 | 52955 |
| eng-ces | newstest2011 | 0.52721 | 25.5 | 3003 | 65653 |
| eng-ces | newstest2012 | 0.50007 | 22.6 | 3003 | 65456 |
| eng-ces | newstest2013 | 0.53643 | 27.4 | 3000 | 57250 |
| eng-ces | newstest2014 | 0.58944 | 31.4 | 3003 | 59902 |
| eng-ces | newstest2015 | 0.55094 | 27.0 | 2656 | 45858 |
| eng-ces | newstest2016 | 0.56864 | 29.9 | 2999 | 56998 |
| eng-ces | newstest2017 | 0.52504 | 24.9 | 3005 | 54361 |
| eng-ces | newstest2018 | 0.52490 | 24.6 | 2983 | 54652 |
| eng-ces | newstest2019 | 0.53994 | 26.4 | 1997 | 43113 |
该工作得到 European Language Grid 的支持,作为 pilot project 2866 的一部分,受到 FoTran project 的支持,该项目由欧洲研究理事会(ERC)在欧洲联盟的Horizon 2020研究和创新计划(授予编号为771113的授权协议)以及欧洲联盟的Horizon 2020研究和创新计划(授权协议编号为780069)下的 MeMAD project 的资助。我们还感谢 CSC -- IT Center for Science 提供的慷慨计算资源和IT基础设施,芬兰。