Joshua Caffery's Team Trains AI on Centuries-Old Louisiana French to Aid 1 Endangered Dialect
Updated
Updated · The New York Times · Jun 27
Joshua Caffery's Team Trains AI on Centuries-Old Louisiana French to Aid 1 Endangered Dialect
1 articles · Updated · The New York Times · Jun 27
Summary
A small University of Louisiana at Lafayette team is building an automatic speech-recognition model for Louisiana French, pushing the endangered oral dialect into digital tools.
Alexa's failure to recognize Cajun musician Dewey Balfa and instead suggesting Dua Lipa crystallized the problem for Professor Joshua Caffery: mainstream AI often misses regional language and culture.
The project draws on historical artifacts and recorded interviews, training the model on snippets such as an old French nursery rhyme to improve recognition of Louisiana French speech.
Caffery, who leads the Center for Louisiana Studies, frames the effort as cultural preservation as much as technology, linking the dialect to Louisiana's history, music and identity.
Will an AI become a language's digital archivist, or can it help create new native speakers?
How can AI capture a language's cultural soul, not just its vocabulary and grammar?
Saving Louisiana French: How the LaFLEUR Project Uses AI and Community to Rescue an Endangered Language
Overview
The LaFLEUR Project, led by Dr. Joshua Caffery at the University of Louisiana at Lafayette's Center for Louisiana Studies, marks a pivotal step in preserving Louisiana French. By partnering with the UL Lafayette Informatics Research Institute and drawing on experts like Rachel L. Doherty, LaFLEUR addresses the urgent need for AI that understands Louisiana French, a language often ignored by commercial AI assistants. Inspired by the failure of mainstream technology to recognize Cajun culture, the project aims to ensure that future AI systems can capture the true 'soul' of Cajun and Creole heritage, safeguarding this unique language for generations to come.