Knowledge Retrieval Revolution: LLM-Based Student Q&A Systems Enhanced by Ontology-Driven Rules

This systematic review examines the integration of Large Language Models (LLMs) and ontology-based rule systems in developing context-aware question-answering (Q&A) platforms for academic environments, a critical advancement in addressing the challenges of personalised information retrieval from dynamic university documents. In discussing the examination of methodologies like Retrieval-Augmented Generation (RAG), vector databases (i.e., Chroma DB, Pinecone), and semantic filtering approaches, the review has documented the elements by which these technologies improve the accuracy, relevance, and adaptability of AI-generated academic assistants.  The synthesis showed that a hybrid architecture, such as RAG and ontology-based filtering, dramatically increases the ability for response personalisation through enforcing batch-specific access rules and modelling hierarchical relationships within the academic domain while simultaneously reducing the hallucinations sometimes seen in outputs from generic LLMs.  Ongoing challenges raised in the review include difficulties in the scale of real-time updates to ontologies, inconsistencies in the standardisation of metadata across university documents, and the lack of evaluation frameworks created for specific domains that would ultimately measure an alignment to the syllabus.  The review also highlighted the need for further innovation in areas such as automated ontology generation, federated learning, which hides user activity while allowing for personalised responses, and collaborative frameworks combining personalisation with university workforces. By synthesising advancements and gaps in the field, this work provides a roadmap for developing robust, context-sensitive Q&A systems that align with the evolving needs of higher education, ultimately fostering more reliable and equitable access to academic knowledge.