Improving efficiency of recursive theory learning

Inductive learning of recursive logical theories from a set of examples is a complex task characterized by three important issues, namely the adoption of a generality order stronger than θ-subsumption, the nonmonotonicity of the consistency property, and the automated discovery of dependencies between target predicates. Solutions implemented in the learning system ATRE are briefly reported in the paper. Moreover, efficiency problems of the learning strategy are illustrated and two caching strategies, one for the clause generation phase and one for the clause evaluation phase, are described. The effectiveness of the proposed caching strategies has been tested on the document processing domain. Experimental results are discussed and conclusions are drawn.