Sustainable multi-model course FAQ & chatbot for NTU Learn V2

Abstract

This project addresses two interconnected challenges in the university learning environment: students' difficulty in accessing course-specific information on NTULearn and professors' struggle to collect timely and meaningful feedback. An LLM-based educational chatbot solution is proposed that serves as both a centralised information resource for students and an incremental feedback collection mechanism for professors. The system leverages Retrieval-Augmented Generation (RAG) techniques to provide accurate, context-aware responses to course-related queries while seamlessly integrating survey questions into chat conversations. A real-time analytics dashboard allows professors to visualise feedback data collected from students. The prototype is implemented using OpenAI's LLM models, Azure cloud services, and Langchain framework. Ultimately, this solution creates a sustainable cycle where improved course-related assistance leads to increased student engagement with the chatbot, which in turn generates more comprehensive feedback for professors, ultimately enhancing the educational experience for all.