Effective top-k computation in retrieving structured documents with term-proximity support