Meta is working with UNESCO on a new plan to improve translation and speech recognition, Techcrunch reports. As part of its Language Technology Partner Program, Meta is looking for partners willing to provide at least 10 hours of transcribed speech recordings, large written texts (over 200 sentences), and sets of translated sentences. Meta’s goal is to focus on “underserved languages in support of UNESCO’s work,” Meta wrote on its blog.
Meta and UNESCO have now signed an agreement with the government of Nunavut, a northern Canadian territory. The goal is to develop translation systems for the Inuit languages used there, Inuktitut and Inuinnaktun. “Our efforts are particularly focused on underserved languages in support of UNESCO’s work for the International Decade of Indigenous Languages,” Meta says.
As part of the program, Meta is releasing an open-source translation benchmark called BOUQuET, a standardized test for evaluating the performance of artificial intelligence models that perform translation. It will consist of sentences “carefully crafted by linguistic experts” and is looking for participants on a special website.
Meta is showing great interest in AI translation of both text and speech, which is a logical step for a company that connects users around the world. Last year, it demonstrated a tool that uses AI to automatically dub videos in other languages with lip-sync, promising that it would first be available to some video creators in English and Spanish in the US. The company has been gradually rolling out its Meta AI assistant around the world, and it is now available in 43 countries and more than a dozen languages.
