Facebook develops multilingual translation tool powered by AI
21 October, 2020
Facebook said it developed the world's first multilingual machine translation tool that may translate between any couple of 100 languages without first translating into English like existing systems do.
Currently, when translating from German to Arabic, the majority of the available English-centric models first translate German to English and then to Arabic, since English training data may be the most widely available.
However, Facebook's new algorithm model M2M-100 directly translates German to Arabic, providing faster results while preserving you see, the context and the meaning of the text.
“For a long time, AI researchers have been working towards creating a single universal model that can understand all languages across different tasks ... an individual model that keeps translations updated and creates new encounters for billions of folks equally,” Angela Fan, research assistant at Facebook AI Research, said.
M2M-100 “brings us closer to this goal”, Ms Fan said. “Breaking language barriers through machine translation is among the most important ways to bring persons together, provide authoritative information on Covid-19 and keep them safe from harmful content.”
The translation model is open sourced and the initial source codes are created freely available. It will facilitate other independent researchers and technology companies to replicate, modify and additional advance the existing multilingual models according with their requirements.
The device learning-based model is trained on practically 2,200 language pairs, almost ten times a lot more than the prior best models that rely only on English language data.
“We will continue to improve our model by incorporating cutting-edge research, exploring methods to deploy machine translation systems responsibly and creating more specialised architectures,” Ms Fan said.
“The research can further advance how our systems understand text for low-resource languages using unlabelled data,” she added.
Traditional machine translation tools require building separate AI models for each and every language and each task.
A number of the advanced multilingual systems can process multiple languages simultaneously, but their accuracy is compromised since they count on English data to bridge the gap between your source and target languages, Facebook said.
“This approach will not scale effectively on Facebook, where persons post content in more than 160 languages across vast amounts of posts. We are in need of one model that may translate any language to raised serve our community … practically two-third which use a language apart from English,” said Ms Fan.
Facebook said it is using various scaling techniques to create translation data sets and create a universal model with 15 billion parameters to reflect a far more various script of languages. It has recently created data sets with 7.5 billion sentences for 100 languages.
“The quantity of data required for training grows quadratically with the number of languages that people support,” Ms Fan said.
Source: www.thenationalnews.com
TAG(s):