The Most-Specific-Generalization Method

Abstract

In this chapter, we develop the Most-Specific-Generalization (MSG) Method, which, within a restricted setting, inductively infers a logic algorithm from examples. This method is part of our tool-box of methods for instantiating the predicate-variables of a schema. First, in Section 10.1, we define the concept of most-specific- generalization. Then, in Section 10.2, we state the objective of the MSG Method, and introduce some other preliminary terminology. We proceed by increasing difficulty, and start with the simplest form the problem can take. Thus, in Section 10.3, we first discuss the case where every example is a ground atom. Then, in Section 10.4, we study the case where non-ground examples, called general examples, are allowed. General examples have disjunctions and existential variables, and are thus different from properties. Future work and related work are discussed in Section 10.5 and Section 10.6, respectively, before drawing some conclusions in Section 10.7.