Meta emphasizes its support for UNESCO's efforts during the International Decade of Indigenous Languages, highlighting its commitment to promoting linguistic diversity.
Leaked documents have revealed ongoing challenges in moderation, particularly with Arabic-language content, underscoring the need for continued improvements in translation and content moderation.
Meta has launched the Language Technology Partner Program in collaboration with UNESCO, aiming to enhance AI-driven speech recognition and translation.
The program encourages contributions from linguistic experts to establish a comprehensive multilingual machine translation standard.
It seeks partners to provide over 10 hours of speech recordings, transcriptions, written texts, and translated sentence datasets, particularly focusing on underserved languages.
Despite these initiatives, Meta has faced criticism regarding its handling of non-English content, especially during the COVID pandemic when misinformation in languages like Italian and Spanish was inadequately addressed.
In response to these criticisms, Meta is actively working to improve its translation and moderation technologies.
Meta's AI assistant has expanded its reach to 43 countries and now supports over a dozen languages, reflecting its commitment to linguistic diversity.
The company is also testing features like automatic translation for Instagram Reels creators, further expanding its language capabilities.
In 2022, Meta introduced the No Language Left Behind (NLLB) project, an open-source machine translation engine that supports many languages.
Additionally, the Meta Massively Multilingual Speech (MMS) project aims to enable audio transcription for over 1,100 languages, with improvements like zero-shot speech recognition introduced recently.
While Meta presents these initiatives as philanthropic, the company stands to gain from improved translation models and enhanced speech recognition capabilities.
