RAG-Assisted Knowledge Graph Construction for Course Recommendation System

Abstract

The study was conducted under a Design Science Research (DSR) methodology to develop a scalable foundation for AI-driven, skill-based course recommendation. For each education program, the taught skills were automatically derived by configuring a Retrieval-Augmented Generation (RAG) pipeline with GPT-4 and grounding it in program descriptions. The generated skills were modeled as nodes and, together with the corresponding program entities, were loaded into a graph database (Neo4j), thereby instantiating a sustainable, domain-specific Knowledge Graph (KG).