Dipartimento di Ingegneria Elettrica, Universita di Palermo,
Viale delle Scienze, 90128, Palermo, Italy
Dipartimento di Informatica, Sistemistica e Telematica, Universita di Genova,
Genova, Italy
Abstract
This paper deals with the handling of uncertainty in rule-based systems. In such systems, knowledge is in general affected by two possible forms of uncertainty: one related to knowledge and the other to observation. An original method is shown based on the Dempster-Shafer theory of evidence, which allows the expert to define interactively, through conventional linguistic terms, so-called atomic knowledge elements, even in cases where such terms are not mutually exclusive. A system may be implemented using the theory of evidence to represent uncertainty: in particular the paper shows that observed facts are conveniently represented by atomic knowledge elements. Simple and expressive examples are illustrated. How to represent a fact is also shown, expressed as a Boolean combination of basic elements, using only conventional linguistic terms and the related simplification algorithm is illustrated.