A re-ranking method based on cloud model

By introducing cloud model, this paper presents a re-ranking method which improves the accuracy of the IR (information retrieval) while recall is preserved. It is rare in traditional Chinese information retrieval to consider uncertainty while calculating the related degree of the query and each document in the result set. This paper researches IR in a perspective of uncertainty by introducing cloud model, measures the relevance between the query and document by the uncertainty degree that using document represents the query, and then re-ranks the result set. Experiments on NTCIR-5 (the 5th NII Test Collection for IR Systems) document collection for SLIR (Single Language IR) show that this method achieves an 18.08% and 26.50% improvement comparing to the initial retrieval method without any re-ranking.