```html

Imagine getting your knuckles rapped every time you used the wrong word. Not the wrong Scrabble play. The wrong language. That's exactly what happened to Louisiana French speakers for decades, and it nearly silenced one of North America's oldest living tongues.

A Language Built Over Centuries, Erased in One Law

For centuries, Louisiana French was the voice of South Louisiana. Not a curiosity. Not a dialect footnote. The predominant language, spoken by ordinary people doing ordinary things. Then in 1921, Louisiana rewrote its state constitution and declared English the primary language.

That's when the knuckle-rapping started.

Kids who spoke Louisiana French in class got punished for it. Literally hit. The message was efficient and brutal: your language is wrong. Speak the right one or hurt. A few generations of that and suddenly you've got a language on life support.

If you've ever felt weirdly protective of a word (we see you, "irregardless" haters), multiply that feeling by a thousand and you start to get what was lost here.

120,000 Speakers and Counting Down

Here's the number that should stop you: in 2023, the Baton Rouge Advocate estimated roughly 120,000 Louisianans still spoke French. That sounds like a lot until you do the math on centuries of suppression, aging populations, and a language officially classified as endangered.

Louisiana French is not thriving. It's surviving. Those are different things.

In 1968, the Council for the Development of French in Louisiana was created specifically to push back, advancing French through education and community programs. That's the kind of institutional effort it takes to keep a language alive when the state spent decades beating it out of people.

The Fiddler Who Kept the Words Alive

Before we get to the AI part, you need to know about Dewey Balfa. Cajun fiddler, singer, and the person credited with popularizing Cajun music. Music was doing what schools refused to: keeping Louisiana French in people's mouths and ears. Songs carry languages across generations in ways that textbooks don't.

That matters for what comes next.

Training a Model to Hear What Knuckles Tried to Silence

Prof. Joshua Caffery directs the Center for Louisiana Studies at the University of Louisiana at Lafayette. He assembled a team there to do something genuinely difficult: train an automatic speech recognition model specifically for Louisiana French.

The model learns from historical artifacts and interviews. Real voices. Real speakers. The kind of linguistic raw material that Christine Mallinson, a professor of language, literacy, and culture at the University of Maryland, Baltimore County, would recognize as irreplaceable.

This is what language preservation looks like in 2024. Not just dictionaries. Actual speech models that can hear the language the way its speakers hear it, built from recordings of people who lived inside it.

One possibility is that tools like this become the foundation for language learning programs that could reach the 120,000 remaining speakers and their kids. The ASR model could power apps, transcription tools, educational resources that make Louisiana French learnable in ways that feel native rather than academic.

Why Word People Should Care

You're here because you love words. Maybe you're obsessive about Wordle. Maybe you argue about whether "qi" counts (it does, 11 points, use it). Either way, you understand that words aren't just game tokens. They carry meaning, history, identity.

Louisiana French carries about 300 years of that. Recipes, songs, jokes, arguments, love stories, all encoded in a linguistic system that a 1921 law and a generation of knuckle-rapping nearly deleted.

The AI project happening in Lafayette right now is a race against time. When the last fluent speakers are gone, no speech recognition model in the world can recreate what they knew. The training data has to come from living voices, recorded now, preserved carefully.

Languages die one speaker at a time. Turns out they can also be saved the same way, one recorded voice at a time.

Source: Languagehat

```