Similarity of query results in similarity-based databases

Abstract

We study estimations of similarity of query results in a similarity-based database model. The model results as extension of Codd's model of data in which domains are additionally equipped with similarity relations and tuples in data tables have ranks indicating degrees to which tuples match similarity-based queries. We present ranked-based and tuple-based similarity of data tables, similarity closures, prove that relational operations in our model preserve similarity, and provide formulas for estimating similarity of query results based on similarity of input data. Most of the proofs are only sketched or omitted because of the limited scope of the paper. © 2011 Springer-Verlag.