Performing deduction from implicit beliefs

Abstract

An implemented knowledge-based system tool, called JOSIE is presented, that allows specialized representations to be efficiently integrated into a uniform framework. As with previous systems of this type, JOSIE allows different kinds of knowledge to be represented differently. Specialized representations are used to improve an application system's efficiency. One way they gain efficiency is by representing some of their beliefs implicitly. The authors have identified two problems with integrating such specialized representations with a forward-directed rule system. The rule invocation problem is the problem of ensuring that rules are triggered from implicit beliefs. The implicit dependency problem is the problem of integrating dependency maintenance, performed by a specialized representation with a general-purpose truth maintenance system. A solution to these problems is presented for a broad class of specialized representations.<>