Comparative Analysis of Deterministic and Graph Neural Network Based RDFS Materialization Methods

Abstract

This thesis compares deterministic Datalog-based and modern deep learning-based methodologies for Resource Description Framework Schema (RDFS) materialization. The research process was meticulously divided into distinct stages. The central focus was to examine the performance of the two methods regarding efficiency and effectiveness. The results indicated that while deep learning approaches, particularly Graph Neural Networks, demonstrated the capability to handle complex graph-structured data, they were considerably slower than their Datalog counterparts. These findings illuminate both methodologies' strengths and limitations, providing crucial insights for future exploration in this domain.