URank: formulation and efficient evaluation of top-k queries in uncertain databases