USING TERM DEPENDENCIES OF A THESAURUS IN THE FUZZY SET MODEL

Abstract

The conventional fuzzy set model has been criticized to generate incorrect ranked output in certain cases due to undesirable properties of MIN and MAX operators. We have defined an operator class called positively compensatory operators giving high retrieval effectiveness and have enhanced the fuzzy set model by using a pair of positively compensatory operators instead of MIN and MAX operators. If a thesaurus is used to represent the documents stored, we can apply a particular methodology to document ranking. In this paper we propose a new thesaurus-based document ranking method called KB-FSM by expanding the enhanced fuzzy set model with term dependence information from a thesaurus. KB-FSM avoids all the problems of the previous methods such as Relevance, R-Distance and K-Distance. We also show through performance comparison that the proposed method provides higher retrieval effectiveness than the others proposed in the past.