On the Difficulty of Token-Level Modeling of Dysfluency and Fluency Shaping Artifacts

This article likely discusses the challenges of using language models to accurately identify and model dysfluencies (like stutters or hesitations) and the artificial patterns that arise when models are trained to improve fluency. The focus is on the token level, meaning the analysis is at the level of individual words or parts of words. The source being ArXiv suggests this is a research paper.