It is not so simple to get an appropriated level of search result among a large number of targets which might be too general or too specific. Hints for the query are valuable for a user to expand or shrink his next search step, if the hints would be shown with their levels compared with the user's original query. This paper proposes a method to extract feature words of the same level as user's query using restricted bootstrap. Examples are shown to demonstrate the effectiveness of the method on tourism blogs. The paper proposes an evaluation measure for the similarity of levels for words based on WordNet. ©2012 IEEE.