Meta has significantly expanded its multilingual AI capabilities, increasing its text-to-speech and speech-to-text technologies from one hundred to over one thousand one hundred languages. By leveraging public religious recordings, the company also improved spoken language identification to cover four thousand languages, representing a fortyfold increase. This expansion ensures broader access to information for users regardless of their preferred language. To support this preservation effort, Meta has open-sourced its models and code, inviting the research community to collaborate on maintaining endangered languages. The initiative aims to democratize technology access, allowing devices to operate in users' native tongues. This collaborative approach fosters global participation in linguistic preservation, turning isolated research into a shared resource for humanity. This development is highly relevant to open data, as it demonstrates how releasing large, diverse datasets can accelerate technological progress for social good. By making linguistic resources openly available, Meta enables researchers to build upon existing work without barriers. It highlights the power of open ecosystems in solving complex problems like language extinction, proving that shared data drives innovation and inclusivity.
Source: 20minutos.esPublished on 2023-05-25