Probably Approximately Correct Completion of Description Logic Knowledge Bases
We propose an approach for approximately completing a TBox w.r.t. a fixed model. By asking implication questions to a domain expert, our method approximates the subsumption relationships that hold in expert’s model and enriches the TBox with the newly discovered relationships between a given set of concept names. Our approach is based on Angluin’s exact learning framework and on the attribute exploration method from Formal Concept Analysis. It brings together the best of both approaches to ask only polynomially many questions to the domain expert.