About the processing of possibilistic and probabilistic queries

Abstract

In this paper, the issue of querying databases that may contain ill-known values represented by disjunctive weighted sets (possibility or probability distributions) is considered. The queries dealt with are of the form: "to what extent is it possible (or probable, depending on the framework considered) that tuple t belongs to the result of query Q", where Q denotes a usual relational query. In the possibilistic database framework, some previous works resulted in the definition of an evaluation method that does not entail computing the different possible worlds of the database. In this paper, we show that this method cannot be used in the probabilistic database framework in general. On the other hand, we describe an alternative evaluation method that is suitable for probabilistic queries when Q complies with certain constraints. © Springer-Verlag Berlin Heidelberg 2005.