Meta Gandeng UNESCO To Expand Language Support To AI Model

Meta Logo (Photo: Unsplash)

JAKARTA - The Meta Fundal AI Research Team (FAIR) is working with the United Nations Educational, Scientific and Cultural Organization (UNESCO), to expand language support to their AI models.

This partnership aims to develop more inclusive AI translation technology and is able to understand various languages and dialects that are not well known to the wider community.

"Developing models that are able to deal with multilingual problems not only encourages the diversity and inclusiveness of languages in the digital world," Meta said in its official blog.

As part of this collaboration, Meta launched the Language Technology Partner Program, which invites global partners to work together to expand open source language technology.

For partners who wish to join, Meta will ask them to donate recordings of speech for more than 10 hours along with its transcription, text written in large quantities (more than 200 sentences) and a series of translated sentences in various languages.

Mitra akan bekerja sama dengan tim Meta untuk membantu mengintegrasikan bahasa-langu tersebut ke dalam model pendentaan aI dan pengian mesin Meta, untuk nantinya dirilis secara gratis bagi masyarakat.

In addition to partner programs, Meta also launched the Tolok Ukur Translation Open Source, a standard test designed to evaluate the performance of AI models in translating various languages.

Meta invites the global community to access this benchmark, which is available in seven languages, as well as contributes the translations they know, to build a more accurate and broad multilanguage machine translation system.

"Today's announcement is part of our long-term commitment to supporting underserved languages," said Meta.