The idea of a private language shared only among close friends or siblings has always held a certain appeal. Now, a new AI model is making this fantasy technically achievable.
Star Trek’s Klingon comes complete with its own dictionary, J.R.R Tolkien invented over a dozen constructed languages for the Lord of the Rings alone, and Dothraki, built for Game of Thrones, carried the grammar and structure of a real spoken language.
Morris Alper, an incoming assistant professor in computer science at the University of Miami and a machine learning and linguistics researcher, has built a model designed to build on this. The tool, he said, could potentially be used to support under-resourced languages and generate content for video games.
“I see it as a kind of creativity, almost an art — imagining different possible realities of ways of communicating,” he said.
The software, called ConlangCrafter and open to anyone, does not spontaneously invent a loose assortment of made-up words. Instead, it creates a consistent set of linguistic rules first — like writing a grammar book — and then generates sentences that adhere to those rules.
Morris Alper, left, incoming assistant professor in the Department of Computer Science at the University of Miami College of Arts and Sciences. At right is an example of a language created by ConlangCrafter. (University of Miami/Courtesy)
Morris designed this tool with co-authors Moran Yanuka, Raja Giryes, and Gašper Beguš.
“People are looking at AI [for] visual creativity like images or drawings, but I immediately thought of AI and linguistic creativity,” Alper said. “When ChatGPT came out, one of the first things I was playing with was whether you could get it to invent a new language.”
One of the other applications of this technology, he said, could be video game design.
“When I was a kid I would play a lot of Pokemon, and I was always frustrated because the world that you’re playing in is limited,” Alper said, noting that the technology could create distinct languages for alien species or other invented cultures.
Inventing new languages, he suggested, could also help preserve real ones — especially those that are endangered and under-resourced.
“A lot of these [AI] systems work well for English or some other languages like Chinese that have a lot of resources poured into them, but in the world there are thousands of languages that are spoken, and a lot of them have very little or no digital resources,” he said. “Some are even in danger of not being spoken anymore, and I’m interested in making technologies that can support these languages.”
By 2100, more than half of the more than 7,000 languages spoken on Earth may disappear, according to University of Connecticut research.
“If we can invent new languages or make things that work with these languages that don’t exist, that might also give us some tools or insights to help us make better tools for small languages that exist but aren’t supported enough,” said Alper, citing the example of less-spoken indigenous American languages.
