Towards More Informative Answers in Terminological Logics

In this paper, the query-answering problem (also called retrieval problem, in terminological logics) is studied and a new framework is proposed. The declarative semantics extends the classical definition of answer in order to refer to some elements that exist in all models although they may not possess a name. With an appropriate representation of these elements, answers are regular languages and are described by regular expressions. We claim that these regular expressions give a better understanding of the knowledge base that is being queried. The operational semantics relies on a finite automaton which is built, independently of the query, from some completions of the knowledge base. Once this automaton has been generated, answers can be efficiently computed. Although sound, the operational semantics is not complete in the general case, but we state sufficient conditions on the knowledge base under which the two semantics are equivalent.