Mitigating Scoring Errors and Compensating for Nonverbal Subtests in Speech-Based Dementia Assessment
Researchers have developed a novel approach to improve the accuracy of speech-based dementia assessments by integrating transcript-derived scores with Whisper embeddings. This method aims to reduce transcription errors and compensate for the omission of nonverbal subtests, such as motor skills, which are crucial for comprehensive cognitive evaluation. The study demonstrates that these fused representations can effectively approximate expert ratings and accurately differentiate between cognitive status groups, even without all subtest data. AI
IMPACT This research could lead to more accessible and accurate early detection of cognitive impairments through AI-powered speech analysis.