Dr. Andreas Steigmiller

Logic-based formalisms are often used for knowledge representation since they allow for flexibly modelling an application domain. In addition, automated reasoning allows for making implicit knowledge explicit, which can, for example, be used for gaining new information or for checking the modelling. Well-known knowledge representation languages, such as the Web Ontology Language (OWL), are based on description logics which have been specifically developed for knowledge representation. More expressive description logics provide more expressive language features to model the application domain in more detail, but they usually also require more sophisticated and often less efficient reasoning procedures.

During my PhD studies, we developed different algorithms and optimisations that make reasoning for many real-world ontologies on standard reasoning tasks more efficient even if the ontologies use features of more expressive description logics. In particular, we developed different caching methods, new absorption algorithms (i.e., rewriting of axioms in the ontology), and coupling techniques. The latter often enable a simple and very efficient saturation of a big part of the knowledge base such that the slower tableau algorithm has only to be used for few more complicated parts.

The aim of the current DFG project is the optimisation of answering conjunctive queries for more expressive description logics. Although conjunctive queries are a foundational element of more comprehensive query languages such as SPARQL, most SPARQL engines do not consider anonymous individuals (together with other more expressive language features) and, therefore, may not determine all possible results. By developing appropriate absorption techniques, we plan to obtain a query answering apporach that is reasonably efficient for real-world ontologies despite considering the features of more expressive description logics.