Meta launches new program to improve speech and translation AI

Meta is launching a new program by partnering with UNESCO that the company has said that the company will help the development of the future publicly available in the future.

Program, the Language Technology Partner ProgramFinding associates who can contribute more than 10 hours to transcription, lots of written texts and “translated into different languages” sentences. According to Meta, the partners will work with the agency’s AI teams to integrate these languages ​​in AI speech recognition and translation models, which will be free if it is up.

The partners still include the Nunavat government of North Canada, a rare populated area. Some of Nunabut’s residents collectively speak the intuits known as Inuqtut.

“Our efforts, especially in support of UNESCO work, focused on the underVard language,” wrote in a blog post provided by Meta TechCrunch. “In the end, our goal is to create an intelligent system that can understand and respond to the needs of complex people regardless of language or cultural background.”

Complement to the new program, Meta said that it was revealing An open source machine translated benchmark To evaluate the effectiveness of the language translation models. The benchmark, composed of sentences made by linguists, supports seven languages ​​and can be accessed from the face of the AI ​​Development platform from the hugs to the face – and it can contribute.

Meta is making both initiatives as public. However, the company may benefit from the upgraded speech recognition and translation models.

The meta continued Extend the number of languages Its AI-driven assistant, meta AI, supports pilot features like automatic translations for creators and pilot features. Last September, Meta announced that it would start examining an equipment to translate the Instagram Relay Voice, Manufacturers to dub their speech and allow it to auto-lip-sinkThe

Meta’s platforms were aimed at the treatment of content content in the language other than English. According to a report, Facebook has left about 70% Italian- and the Spanish language Covid Incorrect data Similar English language is uninterrupted than only 29% of the wrong information. And Leaked documents from the agency Express that the Arabic-language posts are regularly flagged as a disgusting speech.

Meta says it is taking steps to improve its translation and restraint technology.