HSLD: a hybrid similarity measure for linked data resources

Abstract

The web of data is a set of deeply linked resources that can be instantly read and understood by both humans and machines. A vast amount of RDF data has been published in freely accessible and interconnected data sets creating the so-called Linked Open Data cloud. Such a huge amount of data available along with the development of semantic web standards has opened up opportunities for the development of semantic applications. However, most of the semantic recommender systems use only the link structure between resources to calculate the similarity between resources. In this paper we propose HSLD, a hybrid similarity measure for Linked Data that exploits information present in RDF literals besides the links between resources. We evaluate the proposed approach in the context of a LOD-based Recommender System using data from DBpedia. Experiment results indicate that HSLD increases the precision of the recommendations in comparison to pure link-based baseline methods.