Schema discovery in RDF data sources

Abstract

The Web has become a huge information space consisting of interlinked datasets, enabling the design of new applications. The meaningful usage of these datasets is a challenge, as it requires some knowledge about their content such as their types and properties. In this paper, we present an automatic approach for schema discovery in RDF(S)/OWL datasets. We consider a schema as a set of type and link definitions. Our contribution is twofold: (i) generating the types describing a dataset, along with a description for each of them called type profile; (ii) generating the semantic links between types as well as the hierarchical links through the analysis of type profiles. Our approach relies on a density-based clustering algorithm and it does not require any schema-related information in the dataset. We have implemented the proposed algorithms and we present some evaluation results showing the effectiveness of our approach.