Mental Health Detection Via Chatbot

Authors: Professor Roshan U, Aman Kumar Rana, Hemant Kumar Roy, Kushal Malviya, Md Fardwish Ali

Abstract: There has been a significant rise in mental health problems like emotional instability, anxiety, and stress in the college and younger adult populations. Many people struggling with these kinds of issues do not seek help in a timely manner because of stigma, a lack of understanding or knowledge about mental health, and/or the limited availability of professional supports. To address this gap, we developed an artificial intelligence-enabled mental health chatbot that can identify the user's emotional state through their written words and respond accordingly with empathy. The emotional recognition element of our system uses a Support Vector Machine (SVM) emotion classifier and the dialogue generation aspect of our chatbot employs a Seq2Seq network combined with an attention mechanism. We train our SVM emotion classifier on the EmpatheticDialogues dataset and our Seq2Seq model on the CounselChat dataset. The use of attention features results in significant increases in both emotion recognition accuracy and conversation coherence in our comprehensive experimentation. Although our system will never replace a licensed therapist, we believe that it can provide people with an opportunity to express their feelings and gain awareness of their mental health status before seeking professional assistance.