Explanation of Terminological Reasoning A Preliminary Report Stefan Schlobach Language and Inference Technology, ILLC Universiteit van Amsterdam, NL schlobac@science.uva.nl Ronald Cornet Academic Medical Center Universiteit van Amsterdam, NL R.Cornet@amc.uva.nl Abstract This paper describes our current activities to supply extended reasoning sup- port to knowledge engineers who are building terminologies using Description Logics (DL) reasoners. The new services originate in the development of the DICE 1 terminology where the lack of appropriate debugging or explanation fa- cilities hindered a more efficient (and possibly more concise) construction of a corresponding DL TBox. We discuss a number of alternative methods to explain incoherence of TBoxes, unsatisfiability of concepts and concept subsumption. 1 Introduction Developing a terminology is a time-consuming and error-prone process. DICE, a terminology developed at the AMC in Amsterdam for the unambiguous and unified classification of patients in Intensive Care medicine, defines more than 2400 concepts and uses 45 relations. Let us illustrate some of the problems: in a first version of DICE a “brain” was incorrectly specified, among others, as a “central nervous- system” and “body-part” located in the head. This definition is contradictory as nervous-systems and body-parts are declared disjoint in DICE. Fortunately, current Description Logic reasoners, such as RACER [4], can detect this type of inconsistency and the knowledge engineer can identify the cause of the problem. Unfortunately, many other concepts are defined based on the erroneous definition of “brain” forcing each of them to be erroneous as well. In practice, DL reasoners provide lists of hundreds of unsatisfiable concepts for the DICE TBox and the debugging remains a jigsaw to be solved by human experts, with little additional explanation to support this process. The situation is even worse when ontologies without formal semantics are translated into DL terminologies [11, 3]. Cornet and Abu-Hanna, for example, have studied alternative translations, using e.g. rigid translations of slot-fillers in frames into existentially and universally quantified concepts. In some cases this will be too strong 1 DICE stands for “Diagnoses for Intensive Care Evaluation”.