Selective attribute elimination for categorical data specifications

Semantic data specifications are important components of most object oriented software development methodologies. When making a semantic data specification, one has to classify the different kinds of entities that occur in the part of the real world one wants to specify. This classification can be done in two semantically equivalent ways: by putting them in different entity types, or by putting them in one entity type and distinguishing them by means of an attribute. This redundancy, which exists in almost all semantic data specification systems, leads to problems during the view integration process. If two software engineers have chosen to do this classification in a different way, it is hard to integrate their specifications. In this paper we develop and prove correct a transformation which, given a specification that models the real world in one of these ways, computes a new specification that models the real world in the other way. Such a transformation can be very useful during the view integration process.