Detecting semantic violations in generalised classification structures

Abstract

Proposes an approach to the detection of a broad class of semantic violations within general classification structures of information systems. It is shown how constraints upon a classification structure can be defined by a group of exclusion sets which, in turn, imply a simple set of properties to be validated. Frequently incremental updates require that the structure be temporarily invalid. A notion of semantic transaction is introduced allowing incremental updates to be performed atomically with respect to the classification structure. By utilising differential representations of the sets comprising a classification structure constraints can be verified efficiently for a given sequence of updates. The approach is illustrated in the context of an object oriented database system; specifically, we are concerned with systems comprising collections (or in some terminologies, classes) of data and with constraints between these collections. An outline of the implementation of the semantic transaction mechanism is described in this context.<>