The fundamental team of research on META AI (Fair) focuses on the realization of the advanced intelligence of the machines (friend) – IA which can use human reasoning to perform cognitively cognitively tasks, such as translation – and use it to feed the products and innovations that benefit everyone.
Our work with UNESCO to extend the support of poorly served languages in AI models is an essential element of this effort. The development of models capable of working on multilingual problems and in poorly served languages not only promotes linguistic diversity and inclusiveness in the digital world, but also helps us to create smart systems that can adapt to new situations and learn from experience.
Today, we are delighted to share some of our most recent programs, research and models that support this objective and to offer employees opportunities to contribute to IA translation technologies which integrate a wide range of global languages and dialects.
Linguistic technological partners program
We are looking for partners to collaborate with us on the progression and expansion of META’s open source language technologies, including IA translation technologies. Our efforts are particularly focused on poorly served languages, supporting UNESCO’s work as part of the International Decade of Aboriginal Languages.
We are looking for partners who can contribute more than 10 hours of vocal records with transcriptions, large quantities of written text (200+ sentences) and sets of phrases translated into various languages. Partners will work with our teams to help integrate these languages into voice recognition and machine translation models led by AI, which, when published, will be open and freely made available to the community.
As a partner, you will also have access to technical workshops led by our research teams, where you will learn to take advantage of our open source models to create linguistic technologies. We are happy that the government of Nunavut, in Canada, agreed to work with us on this exciting initiative, collaborating to share data in Inuites Inuktitut and Inuinnaqtun languages.
Join our Linguistic technological partners programPlease fill This Interest form.
Open Source Benchmark
In addition to our linguistic partner program, We are launching an open source automatic translation reference, a standard test that will help assess the performance of AI models that carry out a translation. Composed of sentences carefully designed by linguistic experts, we intend to this reference to present the diversity of human language.
We invite you to access the reference, which is available in seven languages, and bring translations that will be made open source and available for others. We aim to build an unprecedented multilingual machine translation reference.
You can access the reference here.
Our commitment to linguistic diversity
Today’s announcements are part of our long-term commitment to support the sub-services languages. In 2022, we published the No Language Left Behind (NLLB) project, a revolutionary Open Source Machine Translation engine which was the first model of translation of the Neuronal Machine for many languages, and laid the basics of research and future development.
We collaborated with UNESCO and an embroidered face to build a Linguistic translator based on NLLBthat we announcement During the United Nations week for the General Assembly last September.
More recently, to support digital empowerment, which is a key thematic field of the global action plan for the international decade of indigenous languageswe introduced tHe met me Massively multilingual speech (MMS) Project, which evolves the audio transcription at more than 1,100 languages. Since then, we have continued to improve and expand its capacities, including the addition of zero-shot vocal recognition in 2024, which allows it to transcribe audio in the languages it has never seen Previously without prior training.
In the end, our goal is to create smart systems that can understand and meet complex human needs, regardless of language or cultural background. While we continue in this direction, we are delighted to improve and expand automatic translation and other linguistic technologies.