Charlotte Latin School is developing an app, currently named Say Again, that leverages AI technology to help world language teachers by evaluating the pronunciation of students.

Tom Dubick, Chair of the Innovation and Design department, explains the inspiration behind the software: “Last year, my wife was filling in as a French teacher. Every Saturday, she spent three hours listening to audio of students and evaluating them. I thought, ‘There must be a better way to do this.’” The logic behind Say Again is that if it makes it faster and easier to provide feedback on pronunciation and fluidity, teachers can do it more often, which helps students.

The app allows students to record themselves reading a short passage in Spanish, French, or another world language and then upload the audio. Within seconds, it produces personalized feedback on their pronunciation, which is reviewed by the teacher and then shared with the speaker in a student-friendly format  — not for a grade, but as a practice tool that helps them spot strengths and weaknesses. It’s being refined and alpha-tested in the Spanish classes of Upper School World Languages Teacher Adrian Singerman and Upper School World Languages Chair Gretchan Frederick, who are exploring how it can best support their daily practices. “I like it for pronunciation,” says Frederick. “We’re still figuring out how best to use it and what training students need to understand the app’s feedback so that it helps them move the needle.”

The software is not built solely on a large language model AI like ChatGPT; instead, it employs an API (application programming interface) called SpeechAce that specializes in fluency and pronunciation. Dubick says that SpeechAce has told him that the company doesn’t know of any other high school using the company’s software the way that Charlotte Latin is, allowing students to receive the fastest possible feedback on a performance exercise in order to help them improve their future.

The school is proceeding carefully. Student audio files are deleted soon after they are processed, and the app is used only as a practice tool. All assessments and graded work continue to rely on the professional judgment of Latin’s world language teachers. But Dubick is already excited about the possible future uses of Say Again, including his idea of maintaining a database of Latin students’ pronunciation habits across the years, helping them to improve their oral fluency. He says, “It’s been very exciting that AI has made it possible to create this app in a short amount of time.”