Towards smarter educational chatbots: a strategy for continuous knowledge augmentation

Abstract

This project focuses on enhancing Ask Narelle, a chatbot used by College of Computing and Data Science students, by integrating continuous knowledge augmentation to improve its ability to handle questions that the chatbot’s existing knowledge base is unable to answer. It introduces a system where unanswered student queries are forwarded to instructors for responses, which are then automatically integrated into the chatbot’s knowledge base. This streamlines the information collection process, reducing instructor workload and improving chatbot accuracy over time. The solution is implemented as a reusable Python package, creating an innovative and adaptive solution that can be applied to other Retrieval-Augmented Generation (RAG) chatbots. The implementation has also been refined and enhanced through valuable insights gathered from student surveys and robust testing, strengthening its effectiveness in diverse domains.