Linked Data and Graph Theory: Gaining Knowledge through the Structure of Heterogeneous Data

Abstract

Linked data has become an integral part of modernizing cultural heritage collections. Similarly, the Rijksmuseum has transformed its digital collection into a linked data format. In partnership with Q42, the museum is developing a search and browse engine that allows both casual and professional users to explore this rich dataset. This thesis explores the intersection of linked data and graph theory to extract knowledge from the dataset's topology. By utilizing community detection algorithms, we identify clusters of similar actors which can be used in a recommendation engine to facilitate users' ability to explore unknown parts of the cultural heritage collection. This thesis proposes several data aggregation methods, several community detection algorithms, and a novel iterative approach to community detection. In our experiments, these techniques are applied to real-world cultural heritage data from the Rijksmuseum. The resulting clusters are evaluated against a validation set and shown to real-world users in a survey. The findings indicate our approach is promising, with the resulting clusters suitable for use in the recommendation engine.