Preserving Strong Equivalence while Forgetting

A variety of proposals for forgetting in logic programs under different semantics have emerged that satisfy differing sets of properties considered desirable. Despite the achieved progress in devising approaches that capture an increasing number of these properties, the idea that the result of forgetting should preserve the meaning of the initial program for the remaining, non-forgotten, atoms, has not yet been captured. In particular, the existing proposals may not preserve dependency relations between such atoms that are given by the structure of the program. In logic programs, these relations are captured by strong equivalence, but, preserving strong equivalence of two different programs while forgetting does not suffice. Rather, strong equivalence relativized to the remaining atoms should be preserved between the original program and the one that results from forgetting. In this paper, we overcome this deficiency by formalizing the property that captures this maintenance of relations while forgetting, and at the same time a general semantic definition for such a forgetting for arbitrary logic programs. Then, we study forgetting for normal programs under the well-founded semantics, and for programs with double negation under the answer set semantics. In both cases, we focus on efficient syntax-based algorithms that only manipulate the rules in which changes are effectively necessary.