Linguistic and Computational Theories of Grammar (VS1) (P01365)

? Credit Points : 20 ? SCQF Level : 11 ? Acronym : INF-P-LCTG-V

The course is an introduction to the contribution of computer science and informatics to linguistic syntax and semantics, and that of linguistic theory to the theory and practice of computational linguistics and cognitive science. The course concentrates on the problem of explaining universal constraints on human grammars, and the integration of syntax, semantics, and probabilistic processing models. in working language processors. It conveys the characteristics of natural grammars, the theoretical devices that have been proposed to handle them, and the implications for theory and practice of language processing.

The course is complementary to courses which deal with parsing or searching for grammatical analyses, such as Introduction to Computational Linguistics and Data Intensive Linguistics, and courses which deal with linguistic and computational semantics, such as Theories of Natural Language Processing, to all of which linkages will be made.

Entry Requirements

? This course is only available to part year visiting students.

? This course is a variant of the following course : P00888

? Pre-requisites : For Informatics PG students only, or by special permission of the School. Knowledge of (or current enrollment in a course on) a high-level programming language (e.g. Lisp, Scheme, Prolog, ML, Python, Java, etc), and willingness to reason with formal systems (e.g. logic, mathematics).

Subject Areas

Home subject area

Language Processing, (School of Informatics, Schedule O)

Delivery Information

? Normal year taken : Postgraduate

? Delivery Period : Not being delivered

? Contact Teaching Time : 3 hour(s) per week for 10 weeks

Summary of Intended Learning Outcomes

Students who complete this course will emerge with an understanding of natural language grammar as a computational system. Knowing upper and lower bounds to the position on the Chomsky hierarchy of expressive power at which natural languages lie, and the nature of the evidence on that establishes these bounds. They will understand:

-the algorithmic expression and limitations of DCG recognisers and parsers for languages of this class.

-how natural language semantics is linked to natural language semantics, and how this linkage is captured in algorithmic terms.

-the role of statistical optimisation in resolving ambiguity and guiding search for such linguistically expressive parsers, and the use of supervised techniques for grammar induction for achieving wide-coverage.

-the relation of all the major linguistic and computational formalisms, including GPSG, ATN, HPSG, LFG, CCG, "Minimalist" TG, and OT to this core computational model.

They will have experienced the cutting edge of research in the theory of grammar and its connection to practical Natural Language Processing Computational Linguistics, and will be aware of the open research questions in the field and the conceptual tools available to address them. This is assessed by written exam.

The students will be able to write hand-built natural language front-end DCG-based interpreters for practical domains in Prolog and by extension other programming languages. They will also know how to induce linguistically expressive grammars and models from corpora for use with more efficient parsing algorithms that will support wide-coverage for such expressive grammars. This is assessed by a carefully graded series of four programming homeworks developing a DCG based parser capturing context-free unification grammar, long-range dependencies, a Montagovian semantics, and a statistical model for an artificial but realistic labeled corpus.