EmoTrack Framework for Depression Tracking from Counseling Transcripts
A new research paper introduces EmoTrack, a framework for predicting PHQ-8 depression scores from text-based counseling transcripts. The study addresses challenges in both single-session and multi-session regimes. The authors created LongCounsel, a multi-session counseling dataset with session-level PHQ-8 supervision, designed to evaluate repeated-session tracking under partial symptom disclosure and cross-session continuity. EmoTrack combines LLM-extracted clinical signals with frozen models to improve robustness. The paper is published on arXiv with ID 2605.22286.
Key facts
- EmoTrack is a PHQ-8 prediction framework for depression tracking from counseling transcripts.
- The study covers both single-session and multi-session regimes.
- LongCounsel is a new multi-session counseling dataset with session-level PHQ-8 supervision.
- The dataset supports evaluation under partial symptom disclosure and cross-session continuity.
- EmoTrack combines LLM-extracted clinical signals with frozen models.
- The paper is available on arXiv with ID 2605.22286.
- Text-based counseling is an important interface for AI mental-health support.
- The research aims to flag sessions requiring timely human review.
Entities
Institutions
- arXiv