Stuttering is a well-known speech disorder with a prevalence of about 5% in children and 1% in adults. Males are significantly more affected than females. The cause of this speech disorder has not yet been conclusively clarified. It is not considered curable, but it is treatable. Therefore, treatment approaches have been established that focus on treating symptoms, such as using speech-assisting devices and rhythmization exercises. Although this often leads to rapid improvements, a lasting therapeutic success cannot be proven. The behavioral therapy method used by our project partner, the Kasseler Stottertherapie, specifically aims at more fluent speech. This is achieved by modifying the entire speech process with a technique known as “fluency shaping.”

In the context of the ENOM project, we want to investigate, to what extent LLMs can be used to model a scenario-driven conversation with an avatar (human-machine dialog) in order to generate randomized but user-centric dialogs.

If interested, please email Korbinian Riedhammer