A Mental Health Chatbot- MindEase

Abstract

Mental health issues are a growing global concern, yet access to timely support remains limited due to stigma, cost, and resource constraints. This paper presents MindEase, an AI-powered mental health chatbot, documenting its evolution from a traditional Feedforward Neural Network (FNN) to an advanced Retrieval-Augmented Generation (RAG) system. We began with a baseline FNN for intent classification, a common chatbot approach. Real-world testing revealed critical limitations: the FNN could only handle narrow, predefined conversation topics, unsuitable for nuanced mental health discussions. This motivated our transition to a RAG system integrating Gemma 2B LLM with a ChromaDB knowledge base covering diverse mental health topics. This paper demonstrates how the RAG approach addresses the conversational breadth and contextual understanding that mental health support demands.