Abstract
University students often face significant barriers to accessing timely mental health support, including stigma, social discomfort, limited counseling sessions, and extended wait times. Recognizing these challenges, this project introduces the Bee-Bestie Enhanced Mental Health Chatbot, an advanced, empathetic conversational agent designed specifically for university student populations. The chatbot aims to bridge the immediate mental health support gap by providing context-aware, real-time interactions enhanced by sentiment and emotion analysis.
This project leverages the fine-tuned GPT-3.5 Turbo API from OpenAI, integrated with sentiment analysis using the Valence Aware Dictionary and sEntiment Reasoner (VADER), and emotion classification employing EmoBERTa. The chatbot dynamically interprets user input, classifies emotional states, and generates empathetic and contextually relevant responses. Conversations are logged and analyzed using an SQLite database, facilitating session continuity and emotional trend tracking. Furthermore, the project incorporates a locally hosted dashboard, allowing visualization of sentiment and emotion data over multiple sessions, assisting students in self-reflection and emotional awareness.
The evaluation results demonstrate the chatbot's effectiveness in maintaining conversation coherence, accurately classifying sentiment and emotions, and providing meaningful emotional support. Additionally, a simulated emergency alert feature is included to demonstrate potential integration with university crisis support systems. This chatbot presents a practical, scalable, and privacy-conscious solution designed to complement existing university mental health services, thereby fostering proactive emotional well-being and encouraging help-seeking behavior among students.