Query Relaxation using Spreading-Activation and SKOS-Ontologies

Abstract

Digital libraries and archives adopting the Linked Open Data (LOD) approach add descriptive metadata to the objects stored in their inventory in order to facilitate searching and sorting. In many cases the available metadata terms are organized as a controlled vocabulary. Technically, the controlled vocabularies are often provided in the form of SKOS ontologies thus allowing to apply semantic web technologies to establish inter-vocabulary relations and ask queries. However, how can a search over the contents of different digital libraries each of which relies on their own vocabulary be performed without sacrificing recall or having to align the underlying ontologies beforehand? In this paper we present an approach based on query relaxation to solving this problem. Considering the graph nature of ontologies for controlled vocabularies we propose to use a spreading activation algorithm to relax and subsequently transform SPARQL queries in a way that makes them suitable for other vocabularies.