DETECTING MENTAL DISORDERS USING EMOTIONAL PATTERNS FROM SOCIAL MEDIA

Abstract

Millions of people around the world suffer from mental disorders that significantly affect their thoughts and behaviors. Detecting these conditions in a timely manner is both challenging and vital, as early intervention can offer critical support before the illness worsens. A promising strategy for early detection involves analyzing how individuals express themselves, particularly through writing and social media activity. This article explores two computational models developed to identify the existence and variations of emotions expressed by users on social media.The analysis is based on two recent public datasets related to eating disorders and depression.  The outcome demonstrate that the proposed emotional representations effectively capture key indicators of these mental health conditions. Furthermore, combining both models improves performance—matching the most effective technique for identifying depression and coming within 1% of the top-performing approach for anorexia. In addition, these models show potential for enhancing the interpretability of the outcomes.