Our Privacy Statement & Cookie Policy

By continuing to browse our site you agree to our use of cookies, revised Privacy Policy and Terms of Use. You can change your cookie settings through your browser.

I agree

A single AI model that can translate 200 languages

CGTN

 /China Media Group
/China Media Group

/China Media Group

A new artificial intelligence model developed by the American company Meta can translate 200 different languages.

Current AI translation models are based on artificial neural networks. These models usually require a large amount of trained data that can be obtained online. However, data for some specific languages cannot be publicly and cheaply obtained. These languages are also called "low-resource languages."

A new cross-language technology developed by a team of staff from Meta allows artificial neural translation models to learn how to use their existing ability to translate high-resource languages to translate low-resource languages. The team developed an online multilingual translation model called NLLB-200, which can translate across 200 different languages. This research was recently published in the British journal Nature.

Researchers said the translation model can help users of low-resource languages use technologies such as the internet, and the model can also be used for education to help such people obtain more books and paper materials. However, the researchers also said that mistranslations of this model are still possible.

Search Trends