Entity Matching in Digital Humanities Knowledge Graphs

Abstract

We propose a method for entity matching that takes into account the characteristic complex properties of decentralized cultural heritage data sources, where multiple data sources may contain duplicates within and between sources. We apply the proposed method to historical data from the Amsterdam City Archives using several clustering algorithms and evaluate the results against a partial ground truth. We also evaluate our method on a semi-synthetic data set for which we have a complete ground truth. The results show that the proposed method for entity matching performs well and is able to handle the complex properties of historical data sources.