Conceptual
Browse Instruct Curate

Speech-Input Length Thresholds for Speaker-Independent Depression Classification

How the quantity of a speaker's recorded speech governs the accuracy of machine-learning models that classify depression from spontaneous spoken responses, independent of the individual speaker. Covers minimum-length thresholds below which classifiers fail, saturation thresholds beyond which additional speech yields diminishing returns, and the finding that eliciting a new response outperforms extending a saturated one.