Mental Stress Detection Using Bidirectional Encoder Representations From Transformers

Abstract

It seems as if people start losing control as they become easily upset, frustrated, and overwhelmed, having problems in resting and quieting their mind, and also feeling bad about themselves, lonely, worthless, and depressed, and avoiding others. If they have experienced the above symptoms, then there is a chance that they are suffering from mental stress. They have to take proper care of their mental health. Stress can be taken care of if it is properly handled and for that detection of stress or the mental state is necessary to provide proper care. The first step in stress detection is sentiment analysis of the users' daily conversations. The authors have proposed an NLP model and have trained it to produce a score for the input ranging between 0 and 1 where 0 is the negative end and 1 is the positive end. The trained model can predict the scores with an accuracy of above 92% on Twitter.