A Pruning Based Approach for Scalable Entity Coreference

Entity coreference is the process to decide which identi-fiers (e.g., person names, locations, ontology instances, etc.) refer to the same real world entity. In the Seman-tic Web, entity coreference can be used to detect equiv-alence relationships between heterogeneous Semantic Web datasets to explicitly link coreferent ontology in-stances via the owl:sameAs property. Due to the large scale of Semantic Web data today, we propose two prun-ing techniques for scalably detecting owl:sameAs links between ontology instances by comparing the similar-ity of their context graphs. First, a sampling based tech-nique is designed to estimate the potential contribution of each RDF node in the context graph and prune in-significant context. Furthermore, a utility function is de-fined to reduce the cost of performing such estimations. We evaluate our pruning techniques on three Semantic Web instance categories. We show that the pruning tech-niques enable the entity coreference system to run 10 to 35 times faster than without them while still maintain-ing comparably good F1-scores.