Intellibot: A Personalized Behavioral Analysis Chatbot Framework Powered by GPT-3

This paper is focused on helping individuals with autism, paranoid personality disorder (PPD), and other mental illnesses practice and improve their communication skills by offering them a friendly, intelligent, and comfortable environment in them to express their needs and overcome their daily communication struggles by practicing. The research aims to construct a personalized Chatbot based on GPT-3 technology to provide professional help to individuals with psychological issues. We proposed implementing an advanced system that chats with the user, detects their emotion, and replies accordingly; it also sees negative thoughts during user sessions and automatically terminates the session while referring the individual to a professional psychiatrist if needed. The paper discusses the whole process of developing the Chatbot from the preprocessing steps for the training data to an overview of GPT-3 and its capabilities. Our top priority is to guarantee the safety and well-being of our users; therefore, the paper addresses the challenges faced during the development of the Chatbot, including the unintentional encouragement of negative thoughts and how we overcame these challenges by implementing proactive measures. Additionally, the paper describes text classification using RoBERTa, voice classification using the "asapp/sew-d-base-plus-400k-ft-ls100h" model, and a Hidden Markov Model to analyze user emotions and provide appropriate responses. Ultimately, the proposed Chatbot aims to create a friendly and comfortable environment for users while offering professional support, enabling fulfilling connections, and improving their quality of life.