MOSS-6: a multi-label dataset and deep learning model for detecting diverse social support-seeking behaviours in online mental health communities

Abstract

Peer-based social support has proven to be an effective approach for enhancing recovery and coping mechanisms among individuals with mental disorders. Early intervention is crucial in treating mental disorders, as delays can lead to chronic conditions requiring sustained management. However, social stigma and limited public funding for mental health services in Korea have severely restricted access, creating a significant gap between the demand for therapeutic interventions and the available resources. The rapid growth of online platforms offers new opportunities to address this gap by providing tailored support through digital communities. Despite the potential of online social support, existing research has primarily focussed on face-to-face interactions or a narrow range of support types, leaving a significant gap in understanding the full spectrum of support-seeking behaviours in digital environments. To address this issue, we present the MOSS-6 dataset, the first Korean dataset specifically developed to comprehensively identify and classify six distinct types of social support-seeking behaviours in online mental health communities. The dataset was meticulously curated and annotated by mental health professionals/clinicians. Additionally, we developed a deep-learning model to accurately classify these support types. This research not only contributes to the understanding of diverse mental health needs within digital platforms but also lays the groundwork for more effective and personalized online mental health support systems. The MOSS-6 dataset is publicly accessible at https://github.com/dxlabskku/MOSS-6.