Meta has announced the launch of the Language Technology Partner Program, a new initiative in collaboration with UNESCO aimed at enhancing AI-driven speech recognition and translation.
As part of the program, Meta is seeking partners capable of contributing over 10 hours of speech recordings with transcriptions, extensive written text, and translated sentence datasets in a wide range of languages. These contributions will facilitate the advancement of sophisticated artificial intelligence models for speech recognition and translation, which Meta intends to make publicly available as open-source upon their completion.
As stated by Meta, designated partners will engage in direct collaboration with Meta’s AI teams to incorporate a variety of languages into AI-driven speech recognition and translation systems. This initiative constitutes a component of Meta’s extensive endeavours to enhance the accessibility and inclusivity of AI language technology on a global scale.
Current partners include the government of Nunavut, a sparsely populated territory located in Northern Canada. Some residents of Nunavut speak Intuit languages collectively known as Inuktut.
“Our efforts are especially focused on underserved languages, in support of UNESCO’s work,” Meta wrote in a blog post. “Ultimately, our goal is to create intelligent systems that can understand and respond to complex human needs, regardless of language or cultural background.”
In addition to the new program, Meta announced the release of an open source machine translation benchmark aimed at assessing the performance of language translation models. The benchmark features sentences meticulously designed by linguists, supports seven languages, and is available for access and contributions on the AI development platform Hugging Face. Meta is presenting both initiatives as charitable efforts. The company is poised to gain from enhanced speech recognition and translation models.
Meta is broadening the range of languages that its AI assistant, Meta AI, can handle, while also testing new features like automatic translation for creators. In September of last year, Meta revealed plans to test a new feature for Instagram Reels that enables creators to translate their voices, offering the ability to dub their speech and achieve automatic lip-syncing.
Meta’s handling of non-English content on its platforms has faced significant scrutiny. A report indicates that Facebook failed to flag nearly 70% of COVID misinformation in Italian and Spanish, while only 29% of comparable misinformation in English went unaddressed. Leaked documents from the company indicate that posts in Arabic are frequently misidentified as hate speech. Meta has said that it’s taking steps to improve its translation and moderation technologies.