사회적 고립 및 우울증 위험군 조기 발견을 위한 AI 기반 감정 분석 및 전문가 연계 플랫폼 개발

초록

To facilitate the early detection and intervention of socially withdrawn individuals and those in mental health risk groups such as hikikomori, we propose a web platform based on AI-based sentiment analysis of users' daily records. The platform addresses the limitations of traditional teleconsultation services, which rely on users actively seeking help, by enabling efficient risk group identification and feedback provision through AI sentiment analysis of casually written daily records. The sentiment analysis model was trained using KLUE-RoBERTa with the Binary Relevance Approach, allowing it to handle sentences containing complex emotions effectively. The platform's effectiveness is evaluated through performance assessments of the emotion analysis model and demonstrations of its practical implementation. By leveraging a highly accessible web-based environment, the platform is expected to contribute significantly to the early detection of socially isolated and depression-prone individuals and to the resolution of mental health issues.

키워드