Search papers, labs, and topics across Lattice.
Technische Hochschule N眉rnberg Georg Simon Ohm
5
0
6
Minor tweaks in evaluation can drastically inflate performance metrics, revealing the fragility of current multimedia event extraction assessments.
Speech-derived features can significantly enhance dementia assessments, even when critical nonverbal tests are omitted.
LLMs can achieve a 35% reduction in dementia assessment errors through structured feature extraction, highlighting their potential in clinical settings.
Wav2Vec 2.0, despite boosting accuracy in speech-based cognitive impairment detection, introduces significant gender and age biases, misclassifying females and younger individuals at a higher rate.
Finally, a large, clinically-validated German speech corpus is available to benchmark and advance speech-based Alzheimer's disease detection.
