NarraSpace: Multimodal Scene-Grounded Evidence for Picture-Description Dementia Classification
Picture-description dementia classification uses a shared visual scene, but a transcript-level prediction alone does not identify the scene evidence underlying its inputs. NarraSpace links words and speech times to named picture elements and retrieves utterances for predefined scene events. Narrative-spatial descriptors summarize coverage and temporal organization. A frozen lan
Reading in Plain English mode — toggle below to switch
Picture-description dementia classification uses a shared visual scene, but a transcript-level prediction alone does not identify the scene evidence underlying its inputs. NarraSpace links words and speech times to named picture elements and retrieves utterances for predefined scene events. Narrative-spatial descriptors summarize coverage and temporal organization. A frozen lan
This paper introduces a novel approach for dementia classification by integrating visual scene evidence with linguistic analysis, moving beyond simple transcript-level predictions. For ML engineers, understanding how to ground language models in specific visual contexts is crucial for developing more interpretable and robust multimodal AI systems, especially in sensitive domains like healthcare.