On the semantics and evaluation of top-k queries in probabilistic databases

Abstract

We formulate three intuitive semantic properties for top-k queries in probabilistic databases, and propose Global-Topk query semantics which satisfies all of them. We provide a dynamic programming algorithm to evaluate top-k queries under Global-Topk in simple probabilistic relations. For general probabilistic relations, we show a polynomial reduction to the simple case. Our analysis shows that the complexity of query evaluation is linear in k and at most quadratic in database size.