The traditional description logics can be used for representing common and individual knowledge about the world (domain of application). Recently description logics have been extended to allow the representation of the knowledge and the beliefs of multiple agents in one knowledge base [3,11]. In MOTEL we formulate knowledge and belief as additional modal operators.

We are using [1] as a base language. That is, we assume three disjoint alphabets, the set of concept names C, the set of role names R, and the set of individual objects O. The set of concept terms (or just concepts) and role terms (or just roles) is inductively defined as follows. Every concept name is a concept term and every role name is a role term. Now assume that C, D are concepts, and R, S are roles. Then and([C,D]), not(C), some(R,C), atleast(n,R), atmost(n,R) are concept terms, and and([R,S]), inverse(R), restrict(R,C) are role terms. The sentences of are divided into terminological sentences and assertional sentences. If C and D are concepts and R and S are roles then defprimconcept(C,D) and defprimrole(R,S) are terminological sentences. If C is a concept, R is a role, and a, b are individual objects then assert_ind(a,C) and assert_ind(a,b,R) are assertional sentences. As in [1] we do not allow terminological cycles.

For the extended language Mod- we assume in addition that we have an alphabet M of modal operator names. Also, there is a distinguished subset A of the individual objects, called the set of agents. We have a distinguished concept name ` all' denoting the set of all agents with which we express mutual belief. The set of concepts and the set of roles of Mod- contains all the concepts and roles of its sublanguage and in addition it contains the concepts b(m,a,C), d(m,a,C), bc(m,C,D), and dc(m,C,D). A modal context is a list of modal operators of the form b(m,a), d(m,a), bc(m,C), and dc(m,C). If L is a modal context, C and D are concepts and R and S are roles then defprimconcept(L,C,D) and defprimrole(L,R,S) are terminological sentences in Mod-. If C is a concept, R is a role, and a, b are individual objects then assert_ind(L,a,C) and assert_ind(L,a,b,R) are assertional sentences in Mod-.

We use a translational approach to provide the usual inference mechanisms, i.e. solving the consistency, the subsumption, the instantiation and the realization problem. Obviously, knowledge bases can be translated into first-order logic theories. There are also well-known relational translation methods for modal logics. In [8] (also available as we have developed an improved translation method for Mod- which provides an elegant translation of knowledge bases into first-order logic theories. In a prototypical implementation, the MOTEL system, we use a Prolog-based system with loop-checking as inference machine.

next up previous
Next: Quantitative Information Up: The MOTEL Knowledge Representation System Previous: A Description Logic for NLP [an error occurred while processing this directive]