Correlation-based Ranking for Relational Databases

Abstract

There has been a great deal of interest in recent years on ranking query results in relational databases. This paper presents a novel method to rank objects (e.g., tuples) by exploiting the correlations among their attribute values. Given a query, each attribute value is assigned a score according to mutual occurrences with the query and its distribution status in the columns of the attribute. These attribute value scores are aggregated to get a final score for an object. Furthermore, a concept vector is proposed to provide a synopsis of the attribute value in a given database. A concept vector is utilized to get similar objects. Experimental results demonstrate the performance of our ranking method, RAVC (Ranking with Attribute Value Correlation), in terms of search quality and efficiency.