RNRank: Network-based ranking on relational tuples

Abstract

Conventional relational top-k queries ignore the inherent referential relationships existing between tuples that can effectively link all tuples of a database together. A relational database can be viewed as a network of tuples connected via foreign keys. With respect to the semantics defined over the foreign keys, the most referenced tuples, therefore, can be regarded as either the most influential, relevant, popular, or authoritative objects stored in a relational database according to its domain semantics. In this paper we propose a novel network-based ranking approach to discover those tuples that are mostly referenced in a relational database as top-k query results. Compared with the conventional relational top-k query processing, our approach can provide information about network structured relational tuples and expand top-k query results as recommendations to users using linkage information in databases. Our experiments on sample relational databases demonstrate the effectiveness and efficiency of our proposed RNRank (Relational Network-based Rank) approach. © Springer International Publishing Switzerland 2013.