Meta Launches Initiative to Advance Speech and Translation AI Technology
Meta has unveiled a new initiative, the Language Technology Partner Program, in collaboration with UNESCO to advance speech recognition and translation AI. This program aims to collect speech recordings, transcriptions, written texts, and translated sentences in a variety of underrepresented languages. The goal is to foster the development of open-source AI models that support diverse languages and address linguistic disparities.
Key Features of the Language Technology Partner Program
Meta is seeking collaborators who can provide:
- 10+ hours of speech recordings with transcriptions
- Substantial written text datasets
- Translated sentence sets in different languages
Collaborators will work alongside Meta’s AI teams to integrate these datasets into cutting-edge AI speech and translation models. These models, once completed, will be made open-source, allowing developers worldwide to contribute and build upon them.
Supporting Underserved Languages
A notable partner in the program is the government of Nunavut, a territory in Northern Canada. This partnership focuses on the preservation and integration of Inuktut, a group of Inuit languages spoken by residents of Nunavut.
Meta emphasized its commitment to underserved languages in a blog post provided to TechCrunch:
“Our efforts are especially focused on underserved languages, in support of UNESCO’s work. Ultimately, our goal is to create intelligent systems that can understand and respond to complex human needs, regardless of language or cultural background.”
Complementary Initiatives
To complement this program, Meta is also launching an open-source machine translation benchmark designed to evaluate language translation models. Created by linguists, this benchmark supports seven languages and is accessible on the AI development platform Hugging Face, enabling researchers and developers to contribute.
Strategic Benefits for Meta
While Meta frames these initiatives as philanthropic, the company is poised to gain from enhanced AI capabilities. The improvements will bolster features such as:
- Expanding the number of languages supported by Meta AI.
- Piloting tools like automatic translation for content creators, including voice translation for Instagram Reels.
Addressing Historical Challenges
Meta has faced criticism for its handling of non-English content on its platforms. Reports have revealed gaps in moderation and translation accuracy, particularly for Arabic, Italian, and Spanish content. The company has acknowledged these issues and claims to be investing in improved technologies to bridge the gaps.
