Interactive multidimensional modeling of linked data for exploratory OLAP. (September 2018)

• Record Type: Journal Article
• Authors: Gallinucci, Enrico; Golfarelli, Matteo; Rizzi, Stefano; Abelló, Alberto; Romero, Oscar
• Is Part Of: Information systems. Volume 77(2018)
• Page Start: 86
• Page End: 104
• Publication Date: 2018-09
• Subjects: Multidimensional modeling -- Data warehouse design -- Linked data -- Exploratory OLAP; Database management -- Periodicals; Electronic data processing -- Periodicals; Bases de données -- Gestion -- Périodiques; Informatique -- Périodiques; Database management; Electronic data processing; Periodicals; 005.7
• DOI: 10.1016/j.is.2018.06.004
• Languages: English
• ISSNs: 0306-4379

Abstract

Highlights: Exploratory OLAP blends OLAP and the Semantic Web to enable cross-domain analyses by adopting a publish-enrich-query paradigm. While some approaches were devised for the public and query stages, the enrich stage has not been investigated yet. We propose the iMOLD approach, that enables data enthusiasts to enrich RDF cubes with aggregation hierarchies by exploring linked data. This is done by detecting five recurring modeling patterns that express roll-up relationships between RDF concepts. A case study based on DBpedia is proposed and the results of an evaluation test made with real users are discussed. Abstract: Exploratory OLAP aims at coupling the precision and detail of corporate data with the information wealth of LOD. While some techniques to create, publish, and query RDF cubes are already available, little has been said about how to contextualize these cubes with situational data in an on-demand fashion. In this paper we describe an approach, called iMOLD, that enables non-technical users to enrich an RDF cube with multidimensional knowledge by discovering aggregation hierarchies in LOD. This is done through a user-guided process that recognizes in the LOD the recurring modeling patterns that express roll-up relationships between RDF concepts, then translates these patterns into aggregation hierarchies to enrich the RDF cube. Two families of aggregation patterns are identified, based on associations and generalization respectively, and the algorithms for recognizing them are described. To evaluate iMOLD in terms of efficiency and effectiveness we compare it with a related approach in the literature, we propose a case study based on DBpedia, and we discuss the results of a test made with real users.