Conceptual Analogy: Conceptual clustering for informed and e cient analogical reasoning

Conceptual analogy (CA) is a general approach that applies conceptual clustering and concept representations to facilitate the e cient use of past experiences (cases) during analogical reasoning (Borner 1995). The approach was developed and implemented in SYN* (see also (Borner 1994, Borner and Fa auer 1995)) to support the design of supply nets in building engineering. This paper sketches the task; it outlines the nearest-neighbor-based agglomerative conceptual clustering applied in organizing large amounts of structured cases into case classes; it provides the concept representation used to characterize case classes and shows the analogous solution of new problems based on the concepts available. However, the main purpose of this paper is to evaluate CA in terms of its reasoning e ciency; its capability to derive solutions that go beyond the cases in the case base but still preserve the quality of cases.