Exploring the Suitability of Transformer Models to Analyse Mental Health Peer Support Forum Data for a Realist Evaluation

Coole, Matthew and Rayson, Paul and Glossop, Zoe and Lobban, Fiona and Marshall, Paul and Vidler, John (2024) Exploring the Suitability of Transformer Models to Analyse Mental Health Peer Support Forum Data for a Realist Evaluation. In: Proceedings of the First Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC-COLING 2024 :. ELRA and ICCL, Torino, Italia, pp. 184-188.

Text (2024.cl4health-1.22)
2024.cl4health-1.22.pdf - Published Version
Available under License Creative Commons Attribution.
Download (264kB)

Abstract

Mental health peer support forums have become widely used in recent years. The emerging mental health crisis and the COVID-19 pandemic have meant that finding a place online for support and advice when dealing with mental health issues is more critical than ever. The need to examine, understand and find ways to improve the support provided by mental health forums is vital in the current climate. As part of this, we present our initial explorations in using modern transformer models to detect four key concepts (connectedness, lived experience, empathy and gratitude), which we believe are essential to understanding how people use mental health forums and will serve as a basis for testing more expansive realise theories about mental health forums in the future. As part of this work, we also replicate previously published results on empathy utilising an existing annotated dataset and test the other concepts on our manually annotated mental health forum posts dataset. These results serve as a basis for future research examining peer support forums.