Nearest Keyword Search on Probabilistic XML Data

Abstract

This paper pays attention to the nearest keyword (NK) problem on probabilistic XML data (NK-P). NK search occupies an important position in information discovery, information extraction and many other areas. Compared with traditional XML data, it is more expensive to answer NK-P search because of so many possible worlds. NK-P can be seen as an NK problem on many traditional XML documents. For a given node q and a keyword k, an NK-P query returns the node which is nearest to q among all the nodes associated with k in all the possible worlds. NK-P search is not only useful independent operator but also as an important part for keyword search. Firstly, we propose a new NK concept on probabilistic XML data based on possible worlds. Next, we present an indexing algorithm to answer an NK-P query efficiently. Finally, extensive experimental results show that our approach is an effective method on probabilistic XML data, and it could significantly reduce the execution time.