Publications / 1987 Proceedings of the 4th ISARC, Haifa, Israel
Knowledge may be conveniently represented in Expert Systems for scheduling constructions by means of frames, or equivalently, by assemblies of feature sets (pairs of the type NAME_OF_FEATURE, VALUE_OF_FEATURE). Each frame is a description of a constraint which is to be observed in a feasible schedule. Such a representation is simple for the user but tends to take sizable memory and extraction of information from it by pattern-matching is time-consuming as well. We define the property of decomposability of the feature sets. It may not be present in all applications, but it definitely exists in the case of construction scheduling. Utilizing decomposability leads to a very sizable reduction of the knowledge base and opens a possibility of efficient scanning of it. In this paper we define decomposability and discuss its application to feature sets. We also show how to compile the feature sets arising in construction scheduling so as to lead to extremely compact rules and efficient scanning.