Search for More Declarativity

Abstract

Abstract. Good tree search algorithms are a key requirement for inference engines of rule languages. As Prolog exemplifies, inference engines based on traditional uninformed search methods with their well-known deficiencies are prone to compromise declarativity, the primary concern of rule languages. The paper presents a new family of uninformed search algorithms that combine the advantages of the traditional ones while avoiding their shortcomings. Moreover, the paper introduces a formal framework based on partial orderings, which allows precise and elegant analysis of such algorithms.

A Generalized QSQR Evaluation Method for Horn Knowledge Bases

“…It also essentially differs from the "subgoal generalization" technique of SLD-ALG [Vieille 1989]. An alternative way is to use D&B-search [Brodt et al 2009]. When the term-depth bound is fixed, our evaluation procedure runs in polynomial time in the size of extensional relations.…”

Generalizing the QSQR Evaluation Method for Horn Knowledge Bases

“…Alternatively, instead of iterative deepening search one can use D&B-search discovered recently in [Brodt et al 2009], which also runs in polynomial space (w.r.t. the maximal depth of visited nodes in the search space) but avoids repetitions.…”