3 - Entity Resolution Models

Abstract

This chapter discusses the three models of entity resolution (ER). The models are complementary in that they address different levels and aspects of the ER process. The model discussed are: Fellegi-Sunter Model, Stanford Entity Resolution, and Algebraic Model. Fellegi-Sunter Model is a methodology for linking equivalent references by direct matching. The Fellegi-Sunter Model provides a specific algorithm for of resolving pairs of references through probabilistic matching. The Stanford Entity Resolution Framework (SERF) defines a generic model of ER in terms of matching and merging operations that act on pairs of references. Unlike Fellegi-Sunter, the SERF model does not define a specific implementation for the match (or merge) operation but instead focuses on methods for resolving a large set of references by the systematic application of the pairwise operations. The Algebraic Model describes ER from an even higher level of abstraction. It focuses on the outcome of the ER process and on metrics for comparing the outcomes of different ER processes acting on the same set of references.