Abstract
This article reviews semantic underspecification, which has emerged over the last three decades as a technique to capture several readings of an ambiguous expression in one single representation by deliberately omitting the differences between the readings in the semantic descriptions. After classifying the kinds of ambiguity to which underspecification can be applied, important properties of underspecification formalisms will be discussed that can be used to distinguish subgroups of these formalisms. The remainder of the article then presents var ious motivations for the use of underspecification, and expounds the derivation and further processing of underspecified semantic representations.