Investigating different Methods for efficient Retrieval of Generalized Cases

In CBR applications, the traditional concept of a case is that of a point in the space spawned by the case’s attributes. Driven by examinations of several new applications, like the retrieval for intellectual properties (IPs), we proposed the concept of generalized cases [BV99, Ber02]. A generalized case (gc) covers not only one point of the attribute space, but a whole subspace of it. It immediately provides solutions to a set of closely related problems rather than to one single problem only. The solutions a gc represents are very close to each other; basically they should be considered as (slight) variations of the same principle solution. In this paper we explore a more formal and systematic view on gcs by using constraints to express the dependencies between several attributes.