Using OpenStreetMap Data to Create Benchmarks for Description Logic Reasoners

ABSTRACT. Engines for query answering over ontological knowledge bases are becoming increasingly popular and important in areas like the Semantic Web or information integration. They are mostly designed to answer queries over ontologies expressed using various Description Logics and in the presence of large amounts of instance data. This computational task, known as ontology-based query answering (OQA), is an important component in the more general area of ontology-based data access. Unfortunately, it has been acknowledged that judging the performance of current OQA reasoners and their underlying algorithms is difficult due to the lack of publicly available benchmarks that consist of large amounts of real-life instance data. In this paper we describe how benchmarks for OQA systems can be created from the publicly available OpenStreetMap (OSM) geospatial data. To this end, we first develop a formalization of OSM and present a rule-based language to specify the rules to extract instance data from OSM data. The declarative nature of the approach allows various variants of a benchmark to be created via small modifications to the rules of the specification. We describe a highly flexible engine to create a benchmark from a given OSM map and a given set of rules and present some evaluation results.