Complement union for data integration

Abstract

A data integration process consists of mapping source data into a target representation (schema mapping [1]), identifying multiple representations of the same real-word object (duplicate detection [2]), and finally combining these representations into a single consistent representation (data fusion [3]). Clearly, as multiple representations of an object are generally not exactly equal, during data fusion, we have to take special care in handling data conflicts. This paper focuses on the definition and implementation of complement union, an operator that defines a new semantics for data fusion.