Smart companion agent for mental well-being through Deep Learning and NLP

Abstract

Mental disorders are an unfortunate reality among the general population nowadays. Conditions like anxiety; depression may seem trivial on the surface but have serious consequences on an individual’s life. These disorders have shown to be detrimental to health and hamper a person’s general well being. In severe cases, if mental disorders go unnoticed and untreated they can cause permanent damage to one’s personality, drive him/her to social isolation and in worst cases compel the person to commit suicide as a means to end their suffering. Therefore, a need for proper detection and awareness of such diseases in a person emerges. Mental disorders may not show physical symptoms in a person but it is possible to find patterns in people with a potential mental disorder and detect them with the help of modern Machine learning techniques. In addition to this, such methods are completely automated and non-invasive; as a result these systems can also help continuously monitor a person’s mental state. We propose a system that can take various physiological signal readings from the human body as a way to predict distress. Upon detecting a user’s distress, the system tries to converse with the user trained by a knowledge base of conversations of people suffering from mental disorders and can interact with the user in a conversation-like interface as a companion. For this we used a system consisting of BioBERT models(separately for questions and answers) and a couple of FCNN layers.