Abstract
Rough Description Logics (DLs) have been studied as a means for representing and reasoning with imprecise knowledge. It has been shown that reasoning in rough DLs can be reduced to reasoning in a classical DL that allows value restrictions, and transitive and inverse roles. This shows that for propositionally closed DLs, the complexity of reasoning is not increased by the inclusion of rough constructors. However, applying such a reduction to rough EL yields an exponential time upper bound. We show that this blow-up in complexity can be avoided, providing a polynomial-time completion-based algorithm for classifying rough EL ontologies.