Abstract: Recently, most of existing algorithms for keywords search always compute the most compact fragments which contain keywords. However, many of these methods do not take the structure-relationship embedded in the XML document into consideration. For the structural query of XML documents,it always has a higher accuracy as it can capture the information searched by users accurately. In this paper, we combine the methods of structural query together with the keyword information retrieval,and infer the needs of users' query through analyzing the relationship between keywords and the structure of the XMI. document. We introduce the thinking of object-oriented and the query of relaxation into keyword query methods,and propose a new architecture of keyword proximity qucry(Rtolrk). Experimental results show that our method can capture the search-intention of users accurately with high recall and precision.

Key words: XML，Keyword, Relaxation query, Object