As often done in node classification on knowledge graphs we use the positive and negative examples to learn concepts in description logics using refinement operators. This way the classifications should not only be accurate but also explainable by looking at the concept, at least for domain experts depending on the ontology.