Linguistic Modifiers: An Overview

Abstract

The power of approximate reasoning is based upon the use of linguistic variables. A variable (e.g. price)is called numeric if its values are numbers (e.g. 5.000 yuan, BEF 100.000, £20.000, 15 Euro,...)and is called linguistic when its values are linguistic terms (e.g. cheap, rather expensive, not expensive but not very cheap either,...). The set of values of a linguistic variable contains at least one primary term (expensive)and most often its antonym or polar opposite (cheap). All the other terms are constructed from these base terms using logical connectives (and, or), negation (not)and linguistic modifiers (rather, fairly, slightly, very,...). The meaning of every term is represented by a Zadeh fuzzy set and can be derived from the fuzzy sets associated with the base terms (i.e. their meaning).

During the last two decades several authors developed techniques for computing the meaning of modified terms. First we will give an overview of these representations of linguistic modifiers and their main properties. Then we will go into their usefulness and state their pro’;s and contra’;s, in a search for the “ideal” representation of some frequently used adverbs like “very” and “more or less”. Finally we shall demonstrate how the best representations can be applied in approximate reasoning with Zadeh’;s compositional rule of inference and the generalized modus ponens.

Martine De Cock is an Aspirant of the Belgian National Fund for Scientific Research. She would like to thank this institution for funding the research, reported on in this article.