Building Partial Domain Theories from Explanations

We propose the use of the explanations provided by a lazy learning method to build a domain theory. Explanations are understood here as generalizations and as a such we interpret them as domain rules in the same sense that eager learning methods do. Differently than domain theories generated by eager learning methods, theories generated from explanations are partial. In this paper the utility of such partial theories according to the domain complexity is discussed. Moreover the equivalence of "global" in front of "partial" domain theories is compared from experimental results.