A feature-based syntax/semantics interface

Abstract

Syntax/semantics interfaces using unification-based or feature-based formalisms are increasingly common in the existing computational linguistics literature. The primary reason for attempting to specify a syntax/semantics interface in feature structures is that it harmonizes so well with the way in which syntax is now normally described; this close harmony means that syntactic and semantic processing (and indeed other processing, see below) can be as tightly coupled as one wishes — indeed, there need not be any fundamental distinction between them at all. In this paper, we first point out several advantages of the unification-based view of the syntax/semantics interface over standard views. These include (i) a more flexible relation to nonsyntactic constraints on semantics, (ii) a characterization of semantic ambiguity, which in turn provides a framework in which to describe disambiguation, and (iii) the opportunity to underspecify meanings in a way difficult to reconcile with other views. The last point is illustrated with an application to the notorious scope ambiguity problem.