Abstract: search method provides users with a friendly way to query XML data, but a users keyword query may often be an imperfect description of their intention. Even when the information need is well described, a search engine may not be able to return the results matching the query as stated. The task of refining the users original query is first defined to achieve better result quality as the problem of keyword query refinement in XML keyword search, and guidelines are designed to decide whether query refinement is necessary. Four refinement operations are defined, namely term deletion, merging, split and substitution. Since there may be more than one query refinement candidates, the definition of refinement cost is proposed, whic,h is used as a measure of semantic distance between the original query and refined query, and also proposed is a dynamic programming solution to compute the refinement cost. In order to achieve the goal of finding the best refined queries and generate their associated results within a one-time node list scan, a stack-based algorithm is proposed, followed by a generalized partition-based optimization, which improves the efficiency a lot. Finally, extensive experiments have been done to show efficiency and effectiveness of the query refinement approach.