AI TrackerFrontier desk · AEST
Back to news
normalaitech

NarraSpace: Multimodal Scene-Grounded Evidence for Picture-Description Dementia Classification

arXiv cs.CL · 21 Oct 2026, 03:55 AEST

NarraSpace introduces a method for classifying dementia through picture descriptions by linking visual scenes to speech data. The approach achieves high accuracy in distinguishing between Alzheimer's patients and healthy controls using a fixed support vector classifier.

Key facts

Read aloud

NarraSpace introduces a method for classifying dementia through picture descriptions by linking visual scenes to speech data. The approach achieves high accuracy in distinguishing between Alzheimer's patients and healthy controls using a fixed support vector classifier. Picture-description dementia classification uses a shared visual scene, but a transcript-level prediction alone does not identify the scene evidence underlying its inputs. On the official forty eight-participant test set of the Alzheimer's Dementia Recognition through Spontaneous Speech benchmark, a fixed support vector classifier achieves eighty dollars and eighty eight cents\%\pm2.fifty six\%$ mean accuracy over fifty seeds. NarraSpace makes the construction of scene-linked classification inputs explicit and inspectable by connecting event-indexed language-model measurements to their source utterances, speech timing, and extraction records.

Sources in this cluster