Physical Address
304 North Cardinal St.
Dorchester Center, MA 02124
Meta launches a new program in partnership with UNESCO to collect speech and copies that the company said will help develop AI available in the future.
The program, and Language technology partner programLooking for collaborators who can contribute more than 10 hours of speech recordings with copies, large quantities of written text, and groups of sentences translated in “various languages”. According to Meta, the partners will work with the company’s AI teams to integrate these languages into the identification forms and translation of AI, which-upon completion-will be open.
Among the partners so far is the Nonavot government, a population density area in northern Canada. Some Nonavot residents speak Intuit languages combined as inuktut.
“Our efforts are particularly focused on disadvantaged languages, to support UNESCO’s works,” Meta wrote in a blog post presented to Techcrunch. “In the end, our goal is to create smart systems that can understand and respond to the complex humanitarian needs, regardless of the language or cultural background.”
Complementary for the new program, Mita said it is launched An open source automatic translation standard To assess the performance of language translation models. The index, which consists of a camel designed by linguists, supports seven languages and can be reached – and contributed to – from the platform of developing the artificial intelligence platform.
Meta is the framing of both initiatives as charitable. But the company will benefit from the models of recognition and translation.
Meta continues in Expand the number of languages The auxiliary with artificial intelligence, Meta AI, supports, and experimental features such as automatic translation of creators. Last September, Mita announced that she would start testing a tool for translating sounds in Instagram rollers, Allow the creators to formulate their speech and an automatic synchronization.
Meta’s content of content in languages other than English through its platforms was a goal for many criticisms. According to one report, Facebook left approximately 70 % of Italian and Spanish information. Unparalleled compared to only 29 % of wrong information in English. and Documents leaked from the company It reveals that the jobs in Arabic are marked regularly, such as hate speech.
Mita said she took steps to improve her translation techniques and moderation.
