The Case of Anorexia and Depression: Detecting Mental Disorders in Social Media Using Emotional Patterns

Abstract

Millions of people worldwide suffer from one or more mental disorders that impair their
thinking and behaviour. A timely detection of these issues is difficult but critical, as it may
allow people to receive treatment before their illness worsens. One option is to monitor how
people express themselves, such as what and how they write, or, going a step further, what
emotions they express in their social media communications. We examine two computational
representations that aim to model the presence and changes of emotions expressed by social
media users in this study. We base our analysis on two recent public data sets for two major
mental disorders: depression and anorexia. The findings indicate that the presence and
variability of emotions captured by the proposed representations allow for the identification
of important information about social media users suffering from depression or anorexia.
Furthermore, combining both representations can improve performance, matching the best
reported approach for depression and falling just 1% short of the top performer for anorexia.
Furthermore, these representations allow for the addition of interpretability to the results.