Keywords filtering over probabilistic XML data

Probabilistic XML data is widely used in many web applications. Recent work has been mostly focused on structured query over probabilistic XML data. A few of work has been done about keyword query. However only the independent and the mutually-exclusive relationship among sibling nodes are discussed. This paper addresses the problem of keyword filtering over probabilistic XML data, and we propose PrXML ^{{exp, ind, mux}} model to represent a more general relationship among XML sibling nodes, for keywords filtering over probabilistic XML data. kdptab is defined as keyword distribution probability table of one subtree. The Dot product, Cartesian product, and addition operation of kdptab are also defined. In PrXML ^{{exp, ind, mux}} model, XML document is scanned bottom-up and achieve keyword filtering based on SLCA semantics efficiently in our method. Finally, the features and efficiency of our method are evaluated with extensive experimental results.