Learning and Reasoning on Knowledge Graphs: From Graph Neural Networks to Large Language Models

This thesis explores how to improve the way computers understand and reason with knowledge stored in large, structured databases called knowledge graphs. It combines two advanced technologies: graph neural networks, which analyze relationships in data, and large language models, which excel at understanding text. The research develops new methods to make these technologies work together, overcoming challenges like limited data and the complexity of real-world knowledge. These advancements can improve applications like medical diagnosis, legal judgements, and product recommendations. This work contributes to making artificial intelligence systems more accurate, interpretable, and reliable in practical scenarios.