System and Processing View in Similarity Assessment

Abstract

Work on similarity can be shown to follow either a system view or a processing view with the former paying more attention to architectures of similiarity assessing systems and the latter concentrating on similarity metrics. As similarity depends on a number of characteristics (e.g. goals, knowledge, context, common features) both view have their own merits when assessing similarity. In this paper, we“ present a framework of multistage similarity assessment that provides a linkage for a modeling of similarity according to the system and processing view. In so doing, the stages of the system can be evaluated according to both the characteristics of similarity being modeled and the errors possibly made.