Previous lecture examined 3 NMR formalisms. Each captured a NMR convention for interpreting a KB by adding formulae to the KB to eliminate the unintended models. Recall that “adding something” enables us to define the correct semantics. In practice we may instead elect not to alter our KB, but rather to achieve the intended interpretation procedurally. CWA Convention: “If I can’t prove it’s true from what’s in my KB then it’s false”, I.e., From KB ?? P(t) infer ? P(t) Result: Assigns true/false to every ground predicate instance in the KB. Minimizes the number of predicates that are true. Applications: completing databases, KBs, ... Syntactic restrictions: only works for Horn clauses (+ a few other special cases) Caveat: Strong form of closure.