NRDR for the acquisition of search knowledge

Abstract

The contribution of this paper is three-fold: It substantially extends Ripple Down Rules, a proven effective method for building large knowledge bases without a knowledge engineer. Furthermore, we propose to develop highly effective heuristics searchers for combinatorial problems by a knowledge acquisition approach to acquire human search knowledge. Finally, our initial experimental results suggest, that this approach may allow experts to stepwise articulate their introspectively inaccessible knowledge.The development of highly effective heuristics for search problems is a difficult and time-consuming task. We present a knowledge acquisition approach to incrementally model expert search processes. Though, experts do not normally have introspective access to that knowledge, their explanations of actual search considerations seems very valuable in constructing a knowledge level model of their search skills.Furthermore, for the basis of our knowledge acquisition approach, we substantially extend Ripple Down Rules [1], a proven effective method for building large knowledge bases without a knowledge engineer: The conditions may involve yet undefined terms which can be incrementally defined during both, the knowledge acquisition as well as the knowledge maintenance process. The resulting framework is called Nested Ripple Down Rules.Our extension greatly enhances the applicability of Ripple Down Rules. Furthermore, for the acquisition of search knowledge, we developed our system SmS1.2 using our new Nested Ripple Down Rules, which has been employed for the acquisition of expert chess knowledge for performing a highly pruned tree search. Our first experimental results in the chess domain are promising for our knowledge acquisition approach to build heuristic searchers which perform a much more restricted tree search than programs like Deep Blue.KeywordsTree SearchKnowledge AcquisitionSearch StateConcept DefinitionKnowledge EngineerThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.