Atypical Speech

Atypical speech research focuses on understanding and improving the processing of speech deviating from typical patterns, encompassing diverse challenges like dysarthria, aphasia, and speech in children. Current research employs machine learning models, including large language models (LLMs) and deep learning architectures, to address these challenges through techniques like personalized automatic speech recognition (ASR) and improved confidence calibration in diagnostic applications. This work is crucial for enhancing healthcare accessibility and improving the accuracy of speech-based diagnostic tools, particularly for individuals with communication disorders. The development of robust and adaptable models is key to overcoming data scarcity and variability inherent in atypical speech datasets.