Novelty in Deductive Databases

In this paper, a notion of novelty of a formula for a concept w.r.t. a deductive database is investigated from a logical point of view. Intuitively, a formula is new for a concept when inserting the formula together with some additional information in the database allows us to infer an instance of the concept (or its negation) from the resulting database, while this proves impossible when only the additional information is inserted. First, this notion is analysed in the framework of first-order Herbrand monotonic databases. A decision procedure is provided, based on a prime implicants characterization. Then, novelty is investigated in the context of non-monotonic—completed—databases. Actually, four types of novelty are put forward to capture various intuitions about the relations between novelty and forms of non-monotonicity. These types of novelty are proved incomparable in the general case and their decision problems are discussed. Finally, possible applications of novelty in the database field are presented.