A re-ranking method based on concept hierarchy using cloud model

This paper implemented the query terms level´s concept extending to acquire the numerical characteristics of the query terms based on cloud model, elevated the concept of the lower query terms level to the higher query level, used the numerical characteristics of the query we gained from the concept level arisen to re-rank documents. This paper researched document re-ranking from the perspective that knowledge has the characteristics of granularity and uncertainty, unlike the previous methods which barely considered uncertainty in document re-ranking. The method we proposed in this paper improved the information retrieval (IR) accuracy while recall is preserved. Experiments on the NTCIR-5 corpus showed that there are 12.20% and 0.38% improvements in the assessments of rigid and relax respectively.