Improving ABox Reasoning for Efficient and Scalable Object Queries over Large Ontologies

Description logics (DLs), as a decidable fragment of first-order logic, are a family of logic based formalisms for knowledge representation and the mathematical underpinning of modern ontology languages, such as OWL and OWL 2. Equipped with well-defined model-theoretical semantics, description logics gain growing applications in knowledge-based systems for representation and reasoning of knowledge of a given domain, covering fields as diverse as biology, health care, decision support, and semantic webs etc. Most of these DL applications involve intensive querying of the underlying knowledge data that requires reasoning over ontologies. Ontology reasoning, however, is a computationally expensive and complex procedure especially for ontologies in an expressive DL, and it may suffer severe intractability problems when ontology ABoxes are extremely large. Alongside the increasing popularity of using DL in data-intensive applications, considerable attention is now shifting to the development of (optimized) algorithms and strategies for scalable ontology reasoning with large ABoxes. In this thesis, we investigate different techniques that can be used to improve ABox reasoning and object queries over large ontologies, including (i) exploring the modularity of an ontology ABox and (ii) developing a revised Most Specific Concept method for efficient instance checking, both of which allow parallel and distributed ontology reasoning and to take advantage of existing parallel-processing frameworks such as MapReduce