In last years, XML has become the standard for information representing, exchanging, and publishing. So, querying XML data is a well-explored topic in both database (DB) and information retrieval (IR) fields, which shows proximity search is well suited to XML documents, which are often modelled as labelled trees. Based on the observations that there are some semantic restrictions in userpsilas query, this paper first analysis query restrictions on elements and their values, and then proposes to approximately extract XML elements by using semantic restrictions.