Towards Learning to Rank in Description Logics