Multimodal depression detection advances with new ranking and temporal models

One paper introduces a fine-grained multimodal model using advantage-weighting ranking to disentangle overlapping depression features. The other explores conversational temporal dynamics between clinician and patient to improve detection.