Abstract: To deal with the problem of information overload for XML data queries, this paper proposed a contextual preferences based XMI_ query results ranking approach. I}he valucbased query predicates user specified arc treated as the context, and then the probabilistic information model is used on the dataset and query history to speculate the user preferences, estimate the corrclations between the specified and unspecified attribute unit values and the relevance between the unspecified attribute unit values and user preferences as well. Next, scoring function of query results is constructed and the ranking score which is computed by the scoring function is used to rank the query results. Results of experiments showed that the ranking approach proposed has high ranking accuracy and can meet the user needs and preferences as well.

Key words: XMI_, Attribute unit, Contextual preference, Probabilistic information retrieval model, Ranking