Processing Vague Queries in Man-Machine Systems: A Fuzzy Set Approach

Abstract

This short communication describes a possible application of fuzzy set theory to information retrieval systems. The aim is to improve the performances of such systems, namely the quality of selection, when all-or-nothing queries are unable to select a sufficiently small subset of items. The main idea is to rank the elements of this selected subset in term of the overall compatibility of each item with a fuzzy description expressing the user's preferences. This description is assumed of a semantic nature, i.e. with numerical features, as opposed to structural or syntactic ones. The latter features may however be used to select a subset of admissible items in a first step. Then, in a second step, the fuzzy description is simply built through a man-machine dialog. The paper focuses on this second step ; it deals with the representation of fuzzy specifications and their combination for the purpose of defining a compatibility index between the query and the objects stored in the data base.