Fact Updates in Logic Databases

A logic database has two components. One comprises the base knowledge, represented explicitly, called the extensional base. The other stores more general knowledge in the form of derivation rules, and is called the intensional base. One may calculate derived facts in advance and store (or materialize) them. Thus, the response time can be expected to speed up. In this paper, we study a new evaluation technique for incrementally maintaining derived facts whenever updates occur on base facts. Our method is based on the seminaive evaluation so it is easy to implement; it is set-oriented, multiple updates may be dealt within a single manipulation; and insertion and deletion are treated uniformly. The method is evaluated analytically as well as experimentally. We indicate the potential application of our incremental method to a computational linguistic problem, i.e., parsing incomplete sentences. Related work in both areas of logic databases and computational linguistics are discussed.