Application of graph grammars to rule-based systems

Abstract

Graph grammars can easily be considered as models for rule-based systems where solving state-space problems essentially requires searching. Since many AI problems are naturally graphical, graph grammars could narrow the usual gap between a problem and its formal specification. In order to be able to solve such problems in practice one must reduce the effort of search. Here, for graph grammar specifications, the idea of precomputing its rules allows to prune the corresponding search-trees safely by explicitly pointing to those subtrees which are contained in others. Moreover, for some rules it becomes possible to use the information of a rule’s former for to predict its later non-applicability, thus avoiding some redundant, expensive applicability tests. The example of solving a domino game based on breadth-first search demonstrates that indeed some remarkable reductions can be obtained.