IJPAM: Volume 44, No. 1 (2008)


Andre de Korvin$^1$, Ping Chen$^2$, Jeong-Mi Yoon$^3$
$^{1,2,3}$Department of Computer and Mathematical Sciences
University of Houston-Downtown
One Main Street - RM. S705, Houston, TX 77002, USA
$^1$e-mail: korvinA@uhd.edu
$^2$e-mail: chenp@uhd.edu
$^3$e-mail: yoonj@uhd.edu

Abstract.In this paper we propose to design a document retrieval system based on the degree of relevance of a set of terms to the documents. The set of relevance of terms has been divided among a number of expert systems. Each system forms its own set of rules based on CART and subsequent pruning of the tree. The method is observed so that each rule pertains to a number of documents whose average relevance is close to 1. These rules are then fuzzified to take care of boundaries separating the rules. We then propose a possible network to implement the document retrieval and the hybrid learning method to train this net. The rules we obtained are of the Sugeno type and the net is an ANFIS construct. Finally we proposed a method to aggregate the decisions of the different expert systems by using a hierarchical structure involving a gating network.

Received: January 26, 2008

AMS Subject Classification: 68P20

Key Words and Phrases: information retrieval, soft computing

Source: International Journal of Pure and Applied Mathematics
ISSN: 1311-8080
Year: 2008
Volume: 44
Issue: 1