Model-preserving transformations of databases for rule-based expert systems with Boolean-valued weights

摘要

Expert systems with databases containing data and implication-like rules are investigated. To each datum or rule, a pair of weight values is ascribed; these values are supposed to be taken from the Boolean algebra generated by an appropriate logical calculus. These weights can be understood as a necessary and a sufficient condition of the validity of the datum or rule in question, known to a particular expert. A system of transformations is proposed, which make the databases easier to apply, conserving, at the same time, their deductive power defined by the corresponding class of models—‘possible worlds’ admitted by the original database.