Probabilistic ranking for relational databases based on correlations

This paper proposes a ranking method to exploit statistical correlations among pairs of attribute values in relational databases. For a given query, the correlations of the query are aggregated with each of the attribute values in a tuple to estimate the relevance of that tuple to the query. We extend Bayesian network models to provide a probabilistic ranking function based on a limited assumption of value independence. Experimental results show that our model improves the retrieval effectiveness on real datasets and has a reasonable query processing time compared to related work. Categories and Subject Descriptors H.3.3 [Information Storage and Retrieval]: Information search and retrieval; H.2.8 [Database Applications]: Miscellaneous General Terms Algorithms, Experimentation Keywords Ranking, Keyword Search over Structured Data, Correlation, Attribute value, Bayesian Networks

• Type: Journal
• Year: 2010
• Where: CIKM
• Authors: Jaehui Park, Sang-goo Lee
• Keywords: Attribute Value, Bayesian Networks, CIKM 2010, Information Technology, Ranking