Modeling and querying probabilistic XML data

Abstract

A survey on modeling and querying probabilistic XML data with focus on the tradeoff between the ability to express real-world probabilistic data and the efficiency of query evaluation is reported. The families PrXML exp and PrXML cie exhibit a clear tradeoff between the efficiency of query evaluation and the ability to model correlation between probabilistic choices. PrXMl can express correlations between distributional nodes by means of shared events variables, though, evaluation of queries with projection is intractable. The PXDB model is completely different in approach and describes correlations in a p-document of PrXML exp in terms of fixed set of constraints rather than many specific dependencies among distributional nodes. The PIXml model describes probabilistic choices similar to PrXML however, PIXml deviates from P-documents in two aspects where, the probability space and possible worlds are represented by directed acyclic graphs and the probabilities of choosing subsets of children's are defined by intervals, rather than exact values.