Interpreting Xml Keyword Query Using Hidden Markov Model

Original scientific paper Keyword search on XML database has attracted a lot of research interests. As XML documents are very different from flat documents, effective search of XML documents needs special considerations. Traditional bag-of-words model does not take the roles of keywords and the relationship between keywords into consideration, and thus is not suited for XML keyword search. In this paper, we present a novel model, called semi-structured keyword query (SSQ), which understands a keyword query in a different way: a keyword query is composed of several query units, where each unit represents query condition. To interpret a keyword query under this model, we take two steps. First, we propose a probabilistic approach based on a Hidden Markov Model for computing the best mapping of the query keywords into the database terms, i.e., elements, attributes and values. Second, we generate SSQs based on the mapping. Experimental results verify the effectiveness of our methods.