Extending SPARQL with Similarity Joins

Abstract

We propose techniques that support the efficient computation of multidimensional similarity joins in an RDF/SPARQL setting, where similarity in an RDF graph is measured with respect to a set of attributes selected in the SPARQL query. While similarity joins have been studied in other contexts, RDF graphs present unique challenges. We discuss how a similarity join operator can be included in the SPARQL language, and investigate ways in which it can be implemented and optimised. We devise experiments to compare three similarity join algorithms over two datasets. Our results reveal that our techniques outperform DBSimJoin: a PostgreSQL extension that supports similarity joins.

Publication
International Semantic Web Conference