Theory-Guideed Induction of Logic Programs by Inference of Regular Languages

Previous resolution-based approaches to theory-guided induction of logic programs produce hypotheses in the form of a set of resolvents of a theory, where the resolvents represent allowed sequences of resolution steps for the initial theory. There are, however, many characterizations of allowed sequences of resolution steps that cannot be expressed by a set of resolvents. One approach to this problem is presented, the system merlin, which is based on an earlier technique for learning nite-state automata that represent allowed sequences of resolution steps. merlin extends the previous technique in three ways: i) negative examples are considered in addition to positive examples, ii) a new strategy for performing generalization is used, and iii) a technique for converting the learned automaton to a logic program is included. Results from experiments are presented in which merlin outperforms both a system using the old strategy for performing generalization, and a traditional covering technique. The latter result can be explained by the limited expressiveness of hypotheses produced by covering and also by the fact that covering needs to produce the correct base clauses for a recursive de nition before producing the recursive clauses. merlin on the other hand does not require that particular examples of the base cases are given, since both base clauses and recursive clauses can be inferred from a single example.