Improving meta speech and translation offers a new program to improve the EU

Meta, a new program launcher with UNESCO, said that the company will openly help the development of the future.

The program, Language technology partner programTranscription, large-sciented text and translated sentences are looking for collaboration seeking cooperation that can contribute to the kits of sentences translated in various languages. According to Methana, partners will work with the EU teams to integrate the EU’s EU speech and translation models to integrate these languages ​​to integrate these languages ​​- when completed.

To date, partners include a sparse settlement area in the North Canada, Canada. Some Nunavut residents, the intuitic languages ​​known as inuktor in collectively.

“Our efforts are focused on the UNESCO’s work, especially in subtle languages, wrote in a blog post given to Meta Techcrunch. “As a result, our goal is to understand intelligent systems that can understand and respond to complex human needs, regardless of language or cultural origin.”

Meta completing the new program said it was released Open source machine Translate Etchark Assess the activities of language translation models. Benchmark, consisting of sentences made by linguists, supported and accessing seven languages ​​and contributes to the AI ​​development platform.

Meta is framed both initiatives as a philanthropist. However, the company stands to take advantage of improved speech recognition and translation models.

Meta continues Expand the number of languages Pilot features like AI-powered assistant, Meta AI, automatic translation for support and creators. Passing September, Meta announced that it will start to test a vehicle to translate sounds in Instagram reels, Allowing the speech of the creators and auto lip synchronization.

Meta was the target of many criticism of other languages ​​from the platforms of contents in other languages ​​in other languages. According to a report, Facebook has left almost 70% Italian and Spanish-speaking Covid error information compared with a total of 29% of similar English-speaking incorrect data. And Leaked documents from the company Arabic-speaking articles are regularly marked as a hatred speech.

He said steps to improve the meta, translation and moderation technologies.