Inferring negative information from disjunctive databases (Q1114440)

Basic concepts of disjunctive databases are defined. Various rules for inferring negative clauses are compared (open and closed world assumption - OWA and CWA, negation as failure, generalized CWA, extended generalized CWA, extended CWA, perfect model rule - PMR). The criteria of comparison are conciseness, efficiency, consistency, decrease, inclusive interpretation of disjunction and (logical) invariability. Two equivalent definitions of a new inference rule, the ``disjunctive database rule'' - DDR - are given. The first of them is based on the concept of a closed set of a database (a subset of the Herbrand base for the first order language of the database). The second definition is based on the least fixed point of a mapping from the lattice of Herbrand interpretations for a database into itself. Properties of the DDR are investigated. The results are extended to the class of layered databases (with predicate hierarchy), where the DDR is implemented as a ``negation as positive failure'' or an ``acyclic negation as positive failure''. The effectivity of implementation and the problem of query evaluation are discussed. With respect to the given criteria, DDR compares favourably with the generalized CWA and PMR. DDR offers a straightforward mechanism for exclusive/inclusive interpretation of disjunction and for predicate protection from inference rule application. Some open problems involving disjunctive databases are reviewed.