In collaboration with UNESCO, Meta is starting a new initiative to gather speech recordings and transcriptions, which the business claims will aid in the creation of future publicly accessible AI.
The Language Technology Partner Program is looking for partners who can provide sets of translated sentences in “diverse languages,” substantial amounts of written content, and over ten hours of speech recordings with transcriptions. In order to incorporate these languages into AI speech recognition and translation models, which will be open-sourced once they are completed, partners will collaborate with Meta’s AI teams, the company said.
The government of Nunavut, a sparsely inhabited territory in Northern Canada, is one of the partners thus far. Inuktut is the collective term for the Intuit languages spoken by some Nunavut natives.
In a blog post given to TechCrunch, Meta stated, “We are particularly focused on underserved languages, in support of UNESCO’s work.” “Our ultimate objective is to develop intelligent systems that, irrespective of language or cultural background, can comprehend and react to complex human needs.”
Complementary to the new program, Meta said that it’s releasing an open source machine translation benchmark to evaluate the performance of language translation models. The benchmark, composed of sentences crafted by linguists, supports seven languages, and can be accessed — and contributed to — from the AI development platform Hugging Face.
ICYMT: Heritage Christian University Lauds UCC for Exemplary Mentorship
Meta is framing both initiatives as philanthropic. But the company stands to benefit from upgraded speech recognition and translation models.
The number of languages supported by Meta AI, its AI-powered assistant, is still growing, and the company is testing capabilities like automatic translation for authors. In September of last year, Meta declared that it will start testing a voice translation tool for Instagram Reels. This feature would enable producers to dub their speech and have it automatically lip-synch.
Meta has come under heavy fire for how it handles information in languages other than English on all of its sites. One investigation claims that Facebook failed to notify nearly 70% of COVID misinformation in Italian and Spanish, whereas just 29% of the same misinformation in English. Additionally, posts in Arabic are frequently mistakenly tagged as hate speech, according to corporate records that have been leaked.
According to Meta, it is working to enhance its moderation and translation systems.
