SEM: a System for Enumerating Models

Model generation can be regarded as a spe­ cial case of the Constraint Satisfaction Problem (CSP). It has many applications in AI, com­ puter science and mathematics. In this paper, we describe SEM, a System for Enumerating fi­ nite Models of first-order many-sorted theories. To the best of our knowledge, SEM outperforms any other finite model generation system on many test problems. The high performance of SEM relies on the following two techniques: (a) an efficient implementation of constraint prop­ agation which requires little dynamic allocation of storage; (b) a powerful heuristic which elim­ inates many isomorphic partial models during the search. We will present the basic algorithm of SEM along with these two techniques. Our experimental results show that general purpose finite model generators are indeed useful in many applications.