Uncertainty Management in Knowledge Modelling

Abstract

Abstract Knowledge base systems have a central role in the architecture of intelligent information systems. The most widely used implementation form is based on ontology modelling. The traditional ontology systems store common objective facts which can be managed with the traditional Boolean logic. The reasoning process is usually based on the strict descriptive logic. In the case of modelling the knowledge database of an agent, the statements are not always strict true or false. There exists some kind of uncertainty in the knowledge model. The paper summarizes the main approaches of uncertainty management and presents a new proposal for a logic model supporting uncertainty in truth value. The classical models assume that the truth value for any predicate can be given with a single value from a scalar value of a lattice domain. In the special case of fuzzy logic, the truth values are coming from the [0,1] interval. These truth values relate to the positive property. On the other hand, we can find in the literature some works arguing for the existence of negative property. In these models, a negative property p- has always a positive counterpart property p+. The proposed logic model is based on the adaption of the infinite-valued probabilistic logic to negative properties. The paper presents a prototype application of the proposed model in the domain of e-tutor systems.