A Neural Network Based on Linguistic Modifiers

Abstract

While the values of a numerical variable are numbers, the values of a linguistic variable are linguistic terms [2]. The term set consisting of all values of a linguistic variable (e.g. Speed) contains one or more base terms: a primary term (e.g. fast), most often also its antonym (e.g. slow), and sometimes a third primary term such as medium. (We want to emphasize that the concepts “primary term” and “antonym” are not interchangeable: the question “how fast is this car?” is neutral, but “how slow is this car?” isn’t.) From these base terms other values can be constructed using conjunction (and), disjunction (or) and modification. In the network model we propose we will only consider base terms which are in fact adjectives, so the only possible modifiers for them are adverbs. We will use the following modifiers: a bit, rather, very, extremely, not ¹.

Martine De Cock and Frederik Vynckier would like to thank the Fund for Scientific Research — Flanders (Belgium) for funding the research reported on in this abstract.

For convenience we consider not as a modifier too, instead of as a logical operator for negation.