Grammaticamodellen in de Taaltechnologie
[ Top | Cursus | Literatuur | Hoorcollege | Werkcollege | Eindopdracht ]

Periode 2: November 2003 - Januari 2004

Docent Paola Monachesi
Paola.Monachesi@let.uu.nl
tel: 030-2536065
Computationele Linguistiek
Trans 10, kamer 2.13
3512 JK Utrecht
Spreekuur: Wednesday 11.00-12.00  


Cursusbeschrijving
[ Top | Cursus | Literatuur | Hoorcollege | Werkcollege | Eindopdracht ]

This course provides an introduction to the techniques and tools that are necessary for the implementation of precise and extensible grammars which are required both in research and in industrial applications. Computational grammars constitute an essential component of text understanding systems, spell-checkers, grammar-checkers, machine translation systems and communication tools for disabled individuals.

The course will combine background lectures with practical exercises on how to formalize linguistic knowledge working mainly within the framework of Head-driven Phrase Structure Grammar (HPSG) (Pollard and Sag 1987, 1994). This framework has the advantage of being well suited both for theoretical linguistic work and for computational implementations.

HPSG can be defined as a constraint-based, lexicalist approach to grammatical theory. It is based on the assumption that human linguistic sentence processing has a powerful lexical basis. It is for this reason that its notion of phrase structure is built around the concept of a lexical head which encodes information that determines crucial grammatical properties of the phrase it projects. Furthermore, HPSG is a surface-based grammar formalism in that it provides a direct characterization of the actual surface order of string elements in a sentence. It can also be viewed as a declarative approach to grammar since the association between strings and informational elements are defined in terms of what associations are possible and not how they are computed. The HPSG informational elements are feature structures, which are partial function from features to their value.

The implementation of HPSG grammar fragments will be conducted within the processing system HDrug, which has been developed at the University of Groningen for formalization and visualisation of constraint-based grammars.

The course will compare HPSG to another grammatical model, that is Tree Adjoining Grammar (TAG).


Skills/Knowledge that the students are going to acquire during the course:

  • general knowledge of the computational properties of HPSG and TAG and of their similarities and differences;
  • general knowledge of the way specific linguistic phenomena are formalized and implemented in unification based grammar models acquired by means of self investigation of already implemented grammar fragments;
  • relation between the two grammar models and their use in the language technology world;
  • hands-on experience with the implementation of a (unification-based) grammar fragment;
  • presentation of papers by means of visual tools;
  • writing of a scientific article.

[ Dutch version ]



Literatuur
[ Top | Cursus | Literatuur | Hoorcollege | Werkcollege | Eindopdracht ]

Books and papers

The books/papers listed below are meant for reference. Only selected chapters will be used during the course.

Abeille' A. 1993. Les Nouvelles Syntaxes, Grammaires d'unification et analyse du francaise. Paris: Armand Colin.
(Introduction contains an overview of the development of unification grammars).

Pollard C. and I. Sag (1987) Information based syntax and semantics. Stanford: CSLI.

Pollard C. and I. Sag (1994) Head-driven Phrase Structure Grammar. Stanford: CSLI .

Sag I. and T. Wasow (1999) Syntactic Theory. A formal introduction. Stanford: CSLI .
(Appendix on the development of Generative Grammar)

Shieber S. (1986) An introduction to unification-based approaches to grammar. Stanford: CSLI.



Relevant Links

Various grammar formalisms- webpages

HPSG - Logical foundations

HPSG Projects

Various grammar implementation tools - webpages

Material for making slides for the presentation



Hoorcollege
[ Top | Cursus | Literatuur | Hoorcollege | Werkcollege | Eindopdracht ]
Weekdag: Datum: Plaats: Tijd:
Tuesday 18-11 until 20-1 BG 048 (Uithof) 11.00-13.00
Thursday 20-11 until 22-1 KNG 66, 005 13.00-15.00

Onderwijsvorm: 4 uur hoorcollege en 2 uur werkcollege

Overzicht:

Week 1: Tuesday 18/11 Introduction to unification grammars. History: unification grammars as a criticism to Transformational grammars and a way to formalize grammars. Development: DCG,FUG,PATR. Different types of formal grammars. Exercises and discussion.
Thursday 20/11 Features to represent linguistic information. Way features are used in various grammar formalisms. Feature theory. Unification. Subsumption. Exercises and discussion.
Week 2: Tuesday 25/11 HPSG - general description. Linguistic signs as typed feature structures. The use of types and inheritance. Exercises and discussion.
Thursday 27/11 Features used to represent linguistic information in HPSG. Different kinds of features: atomic, complex. Lexical entries in HPSG.
Week 3: Tuesday 2/12 Lexical entries in HPSG. The schemas of immediate dominance and the principles that govern the good formation of the phrases.
Thursday 4/12 Lexical entries in HPSG
Week 4: Tuesday 9/12 Generalizations in the lexicon. Morphology in HPSG.
Thursday 11/12 Morphology + determiners.
Week 5: Tuesday 16/12 Dterminers + complementizers + modifiers
Thursday 18/12 Long distance dependencies.
Week 6: Tuesday 6/1 TEST: 10-13 Trans I, B
Proeftentamen
Thursday 8/1 Raising and control.
Week 7: Tuesday 13/1 TAG: general principles.
Thursday 15/1 HPSG vs. TAG: a comparison of the general principles.
Week 8: Tuesday 20/1 Presentation and discussion of HPSG papers.
Thursday 22/1 Presentation and discussion of HPSG papers.


Werkcollege
[ Top | Cursus | Literatuur | Hoorcollege | Werkcollege | Eindopdracht ]
Weekdag:   Plaats: Tijd:
Thursday KNG 80, zaal 012 15.00 - 17.00

In the werkcollege we will have hands-on sessions to get acquainted with the implementation of HPSG grammars. The system HDrug will be adopted to implement the phenomena discussed during the hoorcollege.

Week: Werkgroep: Onderwerp: Te bespreken op werkgroep: Verplichtte opdracht:
Week 1: Thursday 20/11 Introduction to HDrug. How to use the system. Exercises with types and simple grammars. *Opgaven ##* *Opgaven ##*
Week 2: Thursday 27/11 Write a small grammar for dutch with intrasitive, transitive and ditransitive verbs. *Opgaven ##* *Opgaven ##*
Week 3: Thursday 4/12 Add principles, schemas and macros. Analysis of determiners. *Opgaven ##* *Opgaven ##*
Week 4: Thursday 11/12 The analysis of wh-movement *Opgaven ##* *Opgaven ##*
Week 5: Thursday 18/12 TEST *Opgaven ##* *Opgaven ##*
Week 6: Thursday 8/1 Add complement clauses and verb clusters *Opgaven ##* *Opgaven ##*
Week 7: Thursday 15/1 Add complement clauses and verb clusters *Opgaven ##* *Opgaven ##*
Week 8: Thursday 22/1 A large coverage Dutch HPSG grammar. *Opgaven ##* *Opgaven ##*


Eindopdracht
[ Top | Cursus | Literatuur | Hoorcollege | Werkcollege | Eindopdracht ]

The Final Project consists in a paper of max. 4 pages in which the students, working in couples, can choose among one of the following options:

  • discussion of the formalization and/or the implementation of a phenomenon of choice in one of the grammar models presented during the course (on the basis of existing articles/implementations);
  • comparison between the formalization of a phenomenon already analyzed during the course with the formalization of the same phenomenon either in TAG or in the Minimalist Programme or in Categorial Grammar.

Evaluation

  • Test = 30%
  • Eindopdracht = 20%
  • Practicum = 20%
  • Presentatie + class participation = 30%

In order to pass the course you have to have at least a 5.5.
You can find the results of the course here.



Check het rooster op de CKI web-pagina.