The increasing amount of data available on the Internet has made it more important to create efficient IR (Information Retrieval) systems than ever before. However, the execution efficiency of IR tasks under a given scheme is rarely discussed in research. This paper addresses this particular issue by proposing a method that reduces the complexity in both similarity computation and ranking procedures. With the algorithms proposed, the similarity computation may be completed in linear time and the ranking may take near linear time to be finished. Also, it has the potential to improve relevancy analysis in the ranking procedure.