Virtual teaching assistant eduvisor

Abstract

This project aims to develop a Virtual Teaching Assistant (VTA), an AI-powered chatbot designed to enhance student learning and assist educators in managing course-related queries. The VTA not only provides round-the-clock, course-specific support by answering student questions but also encourages deeper engagement with existing course materials and helps identify common learning gaps. By guiding students toward self-directed learning rather than reliance on shortcuts, the VTA fosters a more active and independent approach to education. Additionally, it will offer data-driven insights into areas where students struggle, enabling both students and educators to identify and address these learning gaps effectively. To improve response accuracy, the chatbot leverages Retrieval-Augmented Generation (RAG), which consists of two key components: the retriever, which searches for relevant course materials, and the generator, which formulates responses based on retrieved content. The web application was built using a Model-View-Controller (MVC) architecture, with Streamlit for the frontend and Python (Flask) for the backend. The chatbot integrates Microsoft Authentication (Azure SSO) to ensure secure user access, allowing only verified NTU students and educators to log in. The entire system is deployed on Azure, ensuring scalability and reliability. The project was evaluated by testing both the retriever and generation components of the RAG framework. The retriever was assessed on its ability to fetch relevant course materials, while the generator was evaluated based on the relevance of its responses to queries and the accuracy of its generated answers. Additionally, user testing was conducted among computing students to analyse the chatbot’s response effectiveness, clarity, and speed. Feedback confirmed that the VTA improved student engagement and learning efficiency, while also helping to reduce the burden of repetitive queries for educators. Overall, the VTA successfully enhances student learning while streamlining educator workflows, demonstrating the effectiveness of AI-driven support in education.