Fusion of Overlapped Result Sets

Joaquim Macedo, António Costa and Vasco Freitas

Universidade do Minho
Departamento de Informática
P-4700-320 Braga, Portugal

Tel.: +351 253 604475
Fax.: +351 253 604471
E-mail: {macedo,costa,vf (at) uminho.pt}


Abstract

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

Keywords: Distributed Information Retrieval, Replication, Data Fusion.


4th European Conference on Research and Advanced Technology for Digital Libraries, Lisboa, Portugal, Sep 18-20, 2000