A Deep Learning Approach to Intent-Driven Mental Health Disorder Detection Using RASA and NLP

This research seeks to design an AI chatbot that will be able to determine the mood of a particular user through emotions or conversation in the virtual world since some students, especially within the 14 to 26 age range, feel uncomfortable reporting their problems physically. The Chatbot is designed to detect mental condition or disease, particularly six diseases, Anxiety, Bipolar, Autism, Depression, BPD, and Schizophrenia. Two deep learning-based models have been used to detect mental condition. The dataset was collected from Reddit where six subreddits, including Anxiety, Bipolar, Autism, Depression, BPD and Schizophrenia were present with their corresponding title, text and disease. The Chatbot is designed with the RASA framework for interacting with a human. GPT-3 is used to generate text to continue the conversation further to make the conversation more realistic. RASA framework is used to build the connection between CNN, BERT or LSTM based detection model. For some experimental text generation job has been implemented with GPT-3 open-source, where after detecting the disease, further approaches were taken.