An easy way do it would be to extract the term vectors from the query and then find related documents computing the cosine similarity measure.