OWL Reasoning in the real world: Searching for godot

Abstract

I will provide an overview of many of the use cases that we looked at to apply OWL ABox reasoning in the real world. The fields we covered included (a) healthcare, and life sciences where the plethora of ontologies might be seen as providing a strong use case for OWL reasoning, (b) information retrieval over unstructured text, (c) master data management, which involves reasoning over product and consumer data incorporated from multiple data sources within an enterprise. In each case, we ran into a series of bottlenecks including incorrect modeling of constructs in the ontology, inherent difficulties in scaling OWL reasoning to real world requirements, needing formalisms outside that of OWL, and needing techniques to semi-automate the construction of ontologies. At least in the use cases we had seen, there is a need for performing TBox OWL reasoning over expressive ontologies, but most realistic uses of ABox reasoning have to be relatively simple in terms of expressivity, for practical reasons.