Fusion of overlapped result sets

Abstract

The existence of replicated documents or indexes in dis- tributed information retrieval introduces some level of overlapping be- tween its component databases. In this paper, a new data fusion method taking in account the overlapping information is compared with a conven- tional fusion method. The overall system e_ectiveness evaluation enables preliminary conclusions about the importance of such parameter for data fusion in distributed information retrieval.