Abstract
We introduce a novel knowledge representation scheme where high- level object declarations are translated into a connectionist network. This scheme benefits from the robustness of connectionist representations, without sacrificing the vital advantages provided by a high-level declaration language. In this paper, we introduce our declaration language and explain its translation into connectionist nets. The inherent capability of the system to complete partial and inconsistent declarations is demonstrated by presenting results from experiments performed with our prototype implementation. We also outline the planned integration of the present representation scheme with rule-based systems, endowing a conventional expert system shell with the ability to cope with incomplete and inconsistent information.
Get full access to this article
View all access options for this article.
