Analysis and Evaluation of Synthetic Data Generation in Speech Dysfluency Detection
This addresses data scarcity for clinical speech dysfluency detection, though it is incremental as it builds on existing synthetic data generation methods.
The paper tackles the problem of scarce annotated data for speech dysfluency detection by creating LLM-Dys, a comprehensive synthetic dysfluent speech corpus with 11 dysfluency categories, which improves an end-to-end detection framework to achieve state-of-the-art performance.
Speech dysfluency detection is crucial for clinical diagnosis and language assessment, but existing methods are limited by the scarcity of high-quality annotated data. Although recent advances in TTS model have enabled synthetic dysfluency generation, existing synthetic datasets suffer from unnatural prosody and limited contextual diversity. To address these limitations, we propose LLM-Dys -- the most comprehensive dysfluent speech corpus with LLM-enhanced dysfluency simulation. This dataset captures 11 dysfluency categories spanning both word and phoneme levels. Building upon this resource, we improve an end-to-end dysfluency detection framework. Experimental validation demonstrates state-of-the-art performance. All data, models, and code are open-sourced at https://github.com/Berkeley-Speech-Group/LLM-Dys.