Discourse Coherence and Discourse Structure (P00884)

? Credit Points : 10 ? SCQF Level : 11 ? Acronym : INF-P-DCDS

Discourse structure is a crucial component in both language interpretation and generation. In interpretation, discourse structure provides clues about how people infer content that's linguistically implicit. In generation, discourse structure is needed for planning text which effectively communicates the intended content, and which sounds natural in that it doesn't make every bit of that content linguistically explicit.

But what exactly is discourse structure? Do we need different notions of discourse structure to handle the wide variety of semantic tasks to which it's applied? How do we represent and interpret discourse structure? And in particular, how does discourse structure constrain the interpretation of linguistic elements such as anaphora, tense and ellided VPs? How do we compute the structure of a discourse during language interpretation, and how do we use that structure to infer content that's not linguistically esplicit? And how do we exploit discourse structure during langauge generation? How do the various knowledge resources that people use to compute discourse structure interact? And given that everything from the lexicon to world knowledge and cognitives states provide clues about the structure of a discourse, what impact does this have on the modularity of NLP architectures? What are the limitations of current theories of discourse structure with regard to their application in real systems?

This course will present and critically assess various answers to these questions. The approaches that will be examined range from the logical, through the descriptive, to the computational. We will examine both theory and application. Several guest speakers from within the Division will present their own research and views on discourse structure.

Entry Requirements

? Pre-requisites : Introduction to Computational Linguistics For Informatics PG students only, or by special permission of the School. This course is interdisciplinary in nature, but prerequisites are kept to a mimimum. In the main, we assume knowledge of first order logic.

Subject Areas

Home subject area

Language Processing, (School of Informatics, Schedule O)

Other subject areas

Speech Processing, (School of Informatics, Schedule O)

Delivery Information

? Normal year taken : Postgraduate

? Delivery Period : Not being delivered

? Contact Teaching Time : 2 hour(s) per week for 10 weeks

Summary of Intended Learning Outcomes

By the end of the course, the students should have a general picture of current research in discourse structure within theoretical and computational linguistics, and they should be able to determine the merits of the various theories, and their potential utility in NLP applications.

Knowledge and understanding of linguistic and AI models of discourse interpretation that utilise concepts such as discourse structure and discourse coherence. An appreciation of the importance of discourse structure, as well as its shortcomings. An increased understanding of problems in semantics and pragmatics.

Skills in formal modelling, assessing the literature and linguistic data analysis.

Some elementary skills in understanding algorithms for building discourse interpretation and dialogue systems.

Ability to develop hypotheses and to design methodologies for testing them. Ability to search and read the literature.