Peter Brand and Tsartlip First Nation knowledge keeper J,SINTEN John Elliott, co-creators of FirstVoices, at Lauwelnew Tribal School in Brentwood Bay, B.C. The FirstVoices platform operates in the Indigenous languages technology space.FirstVoices/Supplied
The Lauwelnew Tribal School and what is now known as FirstVoices joined forces more than 20 years ago to bring ethical Indigenous language tools online.
Almost all First Nations languages in B.C. are currently endangered. Dr. Patrick Littell, associate research officer at the federal government’s Digital Technologies Research Centre, whose work revolves around Indigenous language technologies, says there is a lack of learning materials for most of these languages. “Even if there are some text materials for reading and writing practice, some languages often don’t have audio materials for listening and speaking practice.”
The FirstVoices concept was co-created by J,SINTEN John Elliott, a Tsartlip First Nation elder and knowledge keeper, and linguist Peter Brand, alongside other advocates who sought to document and retain ancient Indigenous languages using technology. As that technology advances, Indigenous collaborators and researchers such as Dr. Littell find themselves taking a cautious approach to tools in fields such as generative AI and large language models (LLM) for preservation and revitalization.
“We have to be careful to involve human experts like teachers and elders in these processes to make sure that outputs are appropriate and be mindful that we don’t undermine the important educational goals that we had in the first place in the chase for shiny new technology,” he says.
Seven Indigenous language families and 36 languages are based in B.C., representing more than 50 per cent of Indigenous languages in Canada. FirstVoices includes pages for 33 of the 36. In total, FirstVoices features sections for more than 85 different Indigenous languages across North America and Australia.
FirstVoices and Animikii, both based in B.C. and founded in 2003, are key players operating in the Indigenous language technology space. FirstVoices uses open-source technology, and its data models are available to the public on the GitHub developer platform to support other projects.
The company’s desktop and mobile keyboard apps use Keyman software, which helps developers type in thousands of different languages, and other companies that incorporate Indigenous culture and language into their platforms have since emerged.
While FirstVoices does not use AI, its team is aware of the massive potential of generative AI to support the efforts of Indigenous educators in language retention. It also brings challenges.
Tracey Herbert, CEO of the First Peoples’ Cultural Council, a B.C. Crown corporation and one of the partners of FirstVoices, points out the possibility that synthetic media generators could produce inaccurate language, hindering the progress made by Indigenous educators and collaborators. Ms. Herbert is also concerned about language sovereignty, or a lack of support for protocols around who has the right to use certain languages and the hurdles associated with training a LLM in languages with few written documents.
Dr. Littell of the Digital Technologies Research Centre echoes these concerns, noting that while the tool box of techniques for language retention has grown, his team is taking a cautious approach to embracing it. “One of the risks of the current generation of AI models is hallucination or fabrication, where the model generates realistic sounding but false answers, or even nonsensical or inappropriate answers,” he says. J,SINTEN
He adds he wants to emphasize that Indigenous elders and collaborators are doing the “real work” of language preservation and that technology is merely a tool to advance their efforts. “Technology doesn’t save languages; people do it by speaking to each other, teaching, learning the language, writing stories, and singing songs.”
FirstVoices is also mindful of the potential impacts of AI on language colonization. “Indigenous lands, languages and culture have already been colonized, and AI could be used as a way to colonize Indigenous knowledge and language data further,” Ms. Herbert says. “As technologies evolve, we will continue to communicate transparently about our decisions and actions regarding AI tools.”
One in a regular series of stories. To read more, visit our Indigenous Enterprises section. If you have suggestions for future stories, reach out to IE@globeandmail.com.
Editor’s note: An earlier version of this article inaccurately described the role of Keyman software in the company's desktop and mobile keyboard apps. This version has been corrected. July 29: This article was updated to refer to FirstVoices co-creator J,SINTEN John Elliott by his name in the SENĆOŦEN language.