In this paper, we take the word (or word sequence) that a reader marked in a document as a query and combine global analysis with local analysis to extract the query's context information from the document where query is. By global analysis, we extract keywords from the whole document to reflect the user's research preference. While in the local analysis step, we disambiguate query by extracting keywords from the text that is around the marked query. We use these keywords to reflect the query's context information and suggest interactive query expansion finally.