Abstract
Constraints are important, not only for maintaining data integrity, but also because they capture natural probabilistic dependencies among data items. A probabilistic XML database (PXDB) is the probability subspace comprising the instances of a p-document that satisfy a set of constraints. In contrast to existing models that can express probabilistic dependencies, it is shown that query evaluation is tractable in PXDBs. The problems of sampling and determining well-definedness (i.e., whether the aforesaid subspace is nonempty) are also tractable. Furthermore, queries and constraints can include the aggregate functions count, max, min, and ratio. Finally, this approach can be easily extended to allow a probabilistic interpretation of constraints. © 2009 ACM.