Using specificity to measure referential success in referring expressions with fuzzy properties

Abstract

In the area of linguistic descriptions of data, one of the most important problems is that of referring expression generation. A referring expression is a noun phrase intended to univocally identify a prespecified object. The referential success of a referring expression can be defined as the extent to which the set of objects (possibility distribution) satisfying the referring expression is a singleton containing the intended object. In this paper we provide a formal framework for measuring referential success of referring expressions with fuzzy features. We provide properties to be satisfied by any measure of referential success, and we make a preliminary study of the relationship between referential success and specificity, particularly by proposing a family of measures of referential success based on specificity measures.