NLG-Based Moderator Response Generator to Support Mental Health

Abstract

The global need to effectively address mental health problems and wellbeing is well recognised. Today, online systems are increasingly being viewed as an effective solution for their ability to reach broad populations. As online support groups become popular the workload for human moderators increases. Maintaining quality feedback becomes increasingly challenging as the community grows. Tools that can automatically detect mental health problems from social media posts and then generate smart feedback can greatly reduce human overload. In this paper, we present a system for the automation of interventions using Natural Language Generation (NLG) techniques. In particular, we focus on 'depression' and 'anxiety' related interventions. Psychologists evaluated the quality of the systems' interventions and results were compared against human (i.e. moderator) interventions. Results indicate our intervention system still has a long way to go, but is a step in the right direction as a tool to assist human moderators with their service.