Efficient Axiomatization of OWL 2 EL Ontologies from Data by means of Formal Concept Analysis (Extended Version)

The Web Ontology Language OWL is based on Description Logics and allows representing knowledge in form of ontologies. Such an ontology consists of axioms that describe the underlying domain globally or locally, and is thus divided into a TBox and an ABox. The TBox is used to express rules that all individuals must satisfy as well as to describe the class hierarchy, and in the ABox individuals are assigned to classes or linked to each other by roles. While building an ABox is as easy as collecting data, constructing a TBox is a complex matter. This article is concerned with the latter task and describes an axiomatization approach in the description logic \(\mathcal{EL}\). We specifically show how from input data a TBox can be computed that is complete in the sense that it entails every TBox axiom valid in the data. Moreover, the resulting TBox is guaranteed to be minimal, i.e., the data cannot be completely described with fewer axioms. In addition, an existing TBox can be taken into account and then only axioms are computed that do not already follow from it. The axiomatization method builds upon Formal Concept Analysis. Furthermore, we investigate how it can be implemented in an efficient way and evaluate the prototype on real-world datasets.Author: Francesco Kriegel (Institute of Theoretical Computer Science, Technische Universität Dresden, Dresden, Germany) Mail address: francesco.kriegel@tu-dresden.de Homepage: http://tu-dresden.de/inf/lat/francesco-kriege