Explainable Multimodal Depression Recognition in Clinical Interviews via PHQ-Aligned Symptom Summarization

Explain-MDRC improves depression recognition interpretability by generating structured symptom summaries aligned with PHQ-8 criteria from clinical text. The PhqCML model integrates these summaries with acoustic and facial cues through PHQ-aware contrastive learning. The authors also introduce Explain-DAIC, a new dataset featuring PHQ-8-aligned summary annotations for training.