Home Research Publications Courses Workshops Tutorials Blog


Context Sensitivity in Lexicon and Discourse

Linguistics Institute
Stanford 2007

Date: July 5-26, 2007

Nicholas Asher

James Pustejovsky


Course Abstract

In this course, we focus on the interaction of lexical semantics with discourse semantics. Specifically, we will explore the integration of Generative Lexicon (GL) and Segmented Discourse Representation Theory (SDRT) processes, along with both the problems and advantages that such an integration brings to semantic theory. Both GL and SDRT are reactions to theories of the lexicon and discourse update that fail to account adequately for a wide variety of phenomena having to do with the pragmatics/semantics interface. What earlier theories lack is an account of how the "composition" of new information in context could in fact alter the information as well as the elements in the context, in ways not predictable within a framework countenancing only operations like lambda conversion or merge. GL and SDRT make this the core of their approach to meaning. Broadly speaking, context-sensitive approaches to both lexical composition and discourse interpretation have a common view about meaning, some of the same formal tools, and some of the same problems. In the first part of the course, we concentrate on the theoretical mechanisms of GL and SDRT, and then on their interaction at an analytical level. In the second part of the course, we turn to recent work on the integration of analytical and probabilistic approaches to modeling discourse structure, using the resources outlined in the first part of the course.

Course Areas: Computational Linguistics, Discourse, Empirical Methods, Semantics/Pragmatics

Prerequisites: Basic knowledge of formal compositional semantics operations and linguistic theory; familiarity with dynamic logic, type theory, and probability is desirable.

Readings for the Course