“…between entities), PLATO also enriches the schema by generalizing from the instance level assertions. To explain this step, let C and D be two classes about which we want to find out whether they should be related on the schema level by one of the partonomic relationships R. From the process just described, we obtain a set MR,C,D of instance level assertions of the form R(a, b), where a ∈ C and b ∈ D. 13 We now add schema level axioms according to the following rules: (1) If, for all a ∈ C, there is a b ∈ D with R(a, b) ∈ MR,C,D, then add the axiom C ∃R.D, which can be expressed in OWL/RDF serialization using the owl:someValuesFrom property restriction. (2) If, for all b ∈ D, there is a a ∈ C with R(a, b) ∈ MR,C,D, then add the axiom D ∃R − .C, were R − indicates the inverse (using owl:inverseOf ) property of R. While this approach seems to be rather crude compared to schema learning methods based on inductive paradigms, 14 it already achieves good results, as can be seen from our evaluation in Section 4.3.…”