started · updated
CapisciAMe app aims to improve AI recognition for dysarthria
Davide Mulfari, a computer engineer from the University of Messina, has developed CapisciAMe, a free mobile application designed to help artificial intelligence recognize atypical speech patterns. The app targets individuals with dysarthria, a motor speech disorder that makes spoken language difficult for standard voice assistants, such as Siri or Alexa, to decipher.
By allowing users to record words and phrases directly from their smartphones, the app collects data for fine-tuning AI models. To date, the project has gathered nearly 300 hours of labeled and verified atypical speech, comprising over 350,000 individual recordings from more than 500 contributors. The application utilizes OpenAI’s Whisper and Nvidia’s Parakeet models to facilitate this improvement process.
Developed without direct funding, the project has received technical support from the University of Sydney for computing power and Cineca in Bologna, which provided access to the Leonardo supercomputer. The app has been downloaded over 2,000 times and is gaining traction among speech therapists and professional associations.