Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DiagESC: Dialogue Synthesis for Integrating Depression Diagnosis into Emotional Support Conversation

Seungyeon Seo, Gary Geunbae Lee

TL;DR

This work introduces Diagnostic Emotional Support Conversation (DiagESC) to meld emotional support with proactive depression screening in dialogues. It presents the DESC dataset, synthesized via PHQ-9-based prompts and safeguarded by strict filtering, to enable text-based depression diagnosis in conversation. Expert counselors validate that DESC achieves superior diagnostic ability and high-quality conversational flow compared to baselines. The study demonstrates the feasibility and utility of integrated diagnostic support in mental health chatbots and releases DESC as a resource for further research and development.

Abstract

Dialogue systems for mental health care aim to provide appropriate support to individuals experiencing mental distress. While extensive research has been conducted to deliver adequate emotional support, existing studies cannot identify individuals who require professional medical intervention and cannot offer suitable guidance. We introduce the Diagnostic Emotional Support Conversation task for an advanced mental health management system. We develop the DESC dataset to assess depression symptoms while maintaining user experience by utilizing task-specific utterance generation prompts and a strict filtering algorithm. Evaluations by professional psychological counselors indicate that DESC has a superior ability to diagnose depression than existing data. Additionally, conversational quality evaluation reveals that DESC maintains fluent, consistent, and coherent dialogues.

DiagESC: Dialogue Synthesis for Integrating Depression Diagnosis into Emotional Support Conversation

TL;DR

This work introduces Diagnostic Emotional Support Conversation (DiagESC) to meld emotional support with proactive depression screening in dialogues. It presents the DESC dataset, synthesized via PHQ-9-based prompts and safeguarded by strict filtering, to enable text-based depression diagnosis in conversation. Expert counselors validate that DESC achieves superior diagnostic ability and high-quality conversational flow compared to baselines. The study demonstrates the feasibility and utility of integrated diagnostic support in mental health chatbots and releases DESC as a resource for further research and development.

Abstract

Dialogue systems for mental health care aim to provide appropriate support to individuals experiencing mental distress. While extensive research has been conducted to deliver adequate emotional support, existing studies cannot identify individuals who require professional medical intervention and cannot offer suitable guidance. We introduce the Diagnostic Emotional Support Conversation task for an advanced mental health management system. We develop the DESC dataset to assess depression symptoms while maintaining user experience by utilizing task-specific utterance generation prompts and a strict filtering algorithm. Evaluations by professional psychological counselors indicate that DESC has a superior ability to diagnose depression than existing data. Additionally, conversational quality evaluation reveals that DESC maintains fluent, consistent, and coherent dialogues.
Paper Structure (30 sections, 3 equations, 7 figures, 11 tables)

This paper contains 30 sections, 3 equations, 7 figures, 11 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Supportive Dialogue System
  4. Depression Detection in Conversation
  5. Dialogue Data Synthesis
  6. Problem Formulation
  7. Methodology
  8. Pre-processing
  9. PHQ-9-based Dialogue Generation
  10. Supporter Utterance Generation
  11. Seeker Utterance Generation
  12. Filtering
  13. Diagnostic Dialogue Generation
  14. Post-processing
  15. Experiments
  16. ...and 15 more sections

Figures (7)

  • Figure 1: Part of an example conversation sample from DESC. The left is the seeker's, and the right is the supporter's utterance. We initiate a diagnostic conversation by inserting a diagnostic question (yellow) instead of a specific supporting emotion utterance (gray). At the end of the conversation, appropriate assistance (pink) is provided based on the severity of the depression.
  • Figure 2: The overview of the DESC synthesis process.
  • Figure 3: The overview of seeker utterance generation
  • Figure 4: Distribution of depression severity labels in DESC. Minimal (0-4), Mild (5-9), Moderate (10-14), Moderately severe (15-19), and Severe (20-27).
  • Figure 5: Distribution of evaluated scores for DESC’s fluency, consistency, and coherence.
  • ...and 2 more figures