On the decomposition of tabular knowledge systems

Recently there has been a growing interest in the decomposition of knowledge based systems and decision tables. Much work in this area has adopted an informal approach. In this paper, we first formalize the notion of decomposition, and then we study some interesting classes of decompositions. The proposed classification can be used to formulate design goals to master the decomposition of large decision tables into smaller components. Importantly, carrying out a decomposition eliminates redundant information from the knowledge base, thereby taking away -right from the beginning -a possible source of inconsistency. This, in tum, tenders subsequent verification and validation more smoothly.