Detmar Meurers: Constraint-based Grammar Implementation (795K)

Autumn 2002, 11787-9, Tue & Thu, 10:30-12:48, 029 Derby Hall and Oxley Solarium

I. Course documents (generally in pdf format):

• Syllabus
• Head-Driven Phrase Structure Grammar: An Introduction as Background for Grammar Implementation
  • Part I: General Introduction
  • Motivating different kinds of phrases
  • Part II: Non-finite constructions
  • Part III: Unbounded dependency constructions

• Implementing HPSG Grammars
  • Part I: Background
  • Part II: The Trale System
• From specification to grammar implementation: An example fragment
  • Specification
  • Implementation: signature, theory.pl
• Some other tiny or small example grammars
• Manuals
  • Ale Manual v.3.2.1
  • Trale Manual v.2.1.4 (Draft)

II. Course description

Background

From the linguistic perspective, the development of grammar fragments can be an important means of obtaining feedback on the empirical consequences of a linguistic theory and the compatibility of the various theories which are integrated in the grammar fragment. I would argue that one can go one step further by stating that comprehensive grammar fragments integrating the state-of-the-art of syntactic theorizing are essential for reestablishing the credibility of generative syntax as a science with a measurable criterion for progress.

From the computational perspective, in contrast to the heydays of deep, theory-driven syntactic analysis (where it was viewed as the only viable way to approach natural language) it is now generally recognized that the variety of language processing tasks can and probably should use a variety of approaches, which differ with respect to the kind of linguistic insight they incorporate. While this raises the question what kind of syntactic and semantic representations are needed for what kind of computational task, it seems clear that there is a need at least for syntactic dependency and high-level semantic and discourse representations.

This Seminar

Participating students are expected to independently develop a grammar implementation project, from exploring the theoretical linguistic basis to implementing increasingly complex fragments, testing and documenting them. This will require each enrolled student to invest at least 4 hours per week to work on the individual projects, in addition to the 4 hours of class time. Failure to do so will result in failing the seminar. On the other hand, the seminar is entirely project driven so that there are no other homeworks or seminar papers.

One of the two weekly seminar sessions will consist of taking turns in presenting and discussing the general issues arising in each of the projects such as defining an appropriate signature, lexical specifications, licensing phrases, defining grammatical principles, as well as discussions of practical implementation issues and those arising from the syntactic theories to be implemented. The second weekly meeting is a practical session used to advance the work on each project and discuss issues on an individual basis.

Based on input from potential participants, here are some potential grammar implementation projects that might be tackled:

• A grammar for detecting mistakes of a learner of Spanish
• Continuing work on the Milca adaption of the English Resource Grammar of the Lingo Project at CSLI
• A German fragment using linearization domains
• Ginzburg & Sag
• Sag & Wasow

Preparation to be done before seminar starts