| Abstract | WENDY (Weighted Embedding Network Discovery of Themes) investigates how artificial intelligence can support the creation, organisation, validation and use of educational knowledge structures. Conducted through the University of Westminster’s Students as Researchers Programme, the project brought together academic staff, doctoral researchers, and undergraduate and postgraduate students within a vertically integrated research team. The project addresses the difficulty of manually constructing and maintaining concept maps from heterogeneous teaching materials. It develops a human-centred methodology in which large language models support the extraction of concepts and relationships, embedding-based methods identify broader themes, educational ontologies provide a formal verification framework, and GraphRAG enables natural-language access to structured module knowledge. A conversational-assistant framework is also proposed to support academics in reviewing and refining generated structures and to help students explore concepts, prerequisites and learning pathways. The Level 4 Mathematics for Computing module and its existing SMARTEST representation provide the principal educational context and expert-curated case study. Rather than delivering a fully integrated production system, WENDY develops and evaluates complementary prototypes, design frameworks, and validation procedures to inform future implementation in SMARTEST or similar graph-capable learning environments. Its main contribution is a reusable methodology that combines AI-assisted automation with computational quality assurance and academic oversight, ensuring that generated educational structures remain traceable, reviewable and pedagogically controlled. |
|---|