Characterizing Subsumption

Abstract

As pointed out in Section 3.3, appropriate characterizations of subsumption form the basis for algorithms solving non-standard inferences. The core of every of these characterizations lies in normal forms which provide finite representations of all implications that can be drawn from concept descriptions. Such normal forms can be stated in terms of certain concept descriptions, graphs, or systems of regular languages. Subsumption can then be described using these objects by recursive algorithms, homomorphisms, or language inclusions. Which type of characterization is employed to solve a non-standard inference problem depends on the DL of choice and the inference problem itself. Not all normal forms and characterizations are equally suitable in every situation.