UNESCO Meta Collaboration Improves AI Translation Quality

Meta has joined hands with UNESCO on a new project to improve the translation and speech recognition capabilities of artificial intelligence (AI) models.

The deal is essentially part of Meta's global Language Technology Partner Program. Parent company Facebook is looking for contributors willing to contribute at least 10 hours of voice recordings with captions, long texts (over 200 sentences), and translated sentence sets. The project aims to focus on "lesser-spoken languages, in support of UNESCO's work."

So far, Meta and UNESCO have signed a partnership with the government of Nunavut, a northern Canadian territory. The goal is to develop a translation system for the Inuit languages ​​spoken there, including Inuktitut and Inuinnaqtun. 'We are particularly focused on less-spoken languages, to support UNESCO's work under the International Decade of Indigenous Languages,' Meta said.

Also as part of this program, Meta is releasing an open-source translation benchmark called BOUQuET – a standardized test for evaluating the performance of AI models that perform translations. BOUQuET will include sentences 'carefully constructed by language experts' against which the translation quality of AI models can be evaluated.

Meta has shown a strong interest in AI translation for both text and voice, a logical move for a company that provides platforms that connect users around the world. Last year, Meta introduced an AI tool that automatically dubs Reels into different languages, syncing lip movements, and promised to roll out a beta test for some of its creators' videos in English and Spanish in the U.S. The company has also been gradually expanding its Meta AI assistant globally, now available in 43 countries and more than 10 languages.

Isabella Humphrey

Update 08 February 2025