introduction to grammar engineering with hpsg project ...yzhang/ge-ss08/lecture-01.pdf · lecture...

Grammar engineeringHPSG and DELPH-IN

Structure of the course

Introduction to Grammar Engineering withHPSG

Project Seminar 2008Lecture 1: General Introduction

Yi Zhang

Department of Computational Linguistics & PhoneticsSaarland University

Language Technology LabGerman Research Center for Artificial Intelligence

21st April, 2008

Yi Zhang Grammar Engineering



Introduction to Grammar Engineering with HPSG

Course InfoType: Project SeminarTime: Monday 10:15 – 11:45Location: Conference Room 2.11, Building C 72Homepage: http://www.coli.uni-saarland.de/∼yzhang/ge


http://www.coli.uni-saarland.de/~yzhang/ge



Outline

1 Grammar engineeringWhat?Why?Who?How?Where? When?

2 HPSG and DELPH-IN

3 Structure of the course




What?Why?Who?How?Where? When?

Outline


2 HPSG and DELPH-IN






What Is Grammar Engineering?

GoalTo build

re-usable computational

grammar resources






GoalTo build

re-usable

computational grammar resources






GoalTo build

re-

usable computational grammar resources






GoalTo build re-usable computational grammar resources





Motivation

Linguistically motivated grammars are variable resourcesRelevant to most of the NLP tasksCan be used to test linguistic hypotheses





Application Scenario

Basic tasksParsingGeneration

Real-world applicationsMachine TranslationGrammar CheckingInformation ExtractionQuestion Answering





Linguists and Computer Scientists





Necessary Components

Linguistic TheoryGrammar Engineering PlatformLinguistic ResourcesMethodology





Linguistic Theories

A Solid Linguistic TheoryRigid mathematical foundationTractable computational modelUniversal to different languages





Linguistic Theories of Choice

Tree-adjoining grammar (TAG)Combinatory Categorial Grammar (CCG)Lexical Functional Grammar (LFG)Head-driven Phrase Structure Grammar (HPSG)





Grammar Engineering Platform

Implementation of the formalism (description language)Grammar editorProcessor: parser, generator, etc.Graphical UIOther tools:

Profiling systemTreebanking tools. . .





Linguistic Resources

Corpora / Treebanks / Testsuites / LexicaReference Grammars

Existing grammars for other languages on the sameplatformExisting grammars for the same language on otherplatformsUniversal reusable grammar components





Hand-Crafted Grammars v.s. Treebank-InducedGrammars

Hand-crafted grammarsLong development circleHigh linguistic precision can be achievedGrammar maintenance is difficult(Potentially) suitable for different tasks

Treebank-induced grammarsLarge annotated treebank is requiredSemi-automatic grammar adaptation and extensionMore suitable for parsing than generation





Where? When?

Here! Now!




Outline


2 HPSG and DELPH-IN





HPSG as a linguistic theory

Highly consistent and powerful formalism (TFS)Monostratal, declarative, non-derivational, lexicalist,constraint-basedHas been studied for many different languagesPsycholinguistic evidence




Mathematical foundation

Typed Feature Structure[Carpenter, 1992]High expressive powerParsing complexity: exponential (to the input length)

Tractable with efficient parsing algorithmsEfficiency can be improved with a well designed grammar




Implementations

Mismatch exists between theory and realityTrade-off has to be made for various practical reasonsWhat makes a good implementation?

Faithfulness to the linguistic theoryReliability of the software toolchainExtensibility




DELPH-INDeep Linguistic Processing with HPSG – Initiative

LKB

Grammar

Generator Parser

TDL

[incr tsdb()]

Heart-of-Gold

Matrix

PET

MRSHPSG

ERG JaCY GG




A glance over existing grammar engineering platforms

TAGXTAG

LFGXLE

HPSGEnju (English)Alpino/HDRUG (Dutch)TRALE




Outline


2 HPSG and DELPH-IN





Things to master

HPSG

MRS

LKB

[incr tsdb()]




Part I: LKB, TFS and TDL

Hands-on introduction to LKB [Copestake, 2002]Type description language




Part II: Minimal Recursion Semantics (MRS)

[Copestake et al., 2005]Semantically informed grammar engineeringCompositional semantics




Part III: Matrix

[Bender et al., 2002]Shared components of a HPSG across different languagesImplemented language phenomena groups as libraries andplug-ins




Part IV: Implementations of various languagephenomena

SubcategorizationModification and agreementsLong distance dependenciesMore




Questions?


References

References I

Bender, E., Flickinger, D., and Oepen, S. (2002).The Grammar Matrix: an open-source starter-kit for the rapid development of cross-linguistically consistentbroad-coverage precision grammars.In Proceedings of the Workshop on Grammar Engineering and Evaluation at the 19th InternationalConference on Computational Linguistics, pages 8–14, Taipei, Taiwan.

Carpenter, B. (1992).The Logic of Typed Feature Structures.Cambridge University Press, Cambridge, UK.

Copestake, A. (2002).Implementing Typed Feature Structure Grammars.CSLI, Stanford, USA.

Copestake, A., Flickinger, D., Pollard, C. J., and Sag, I. A. (2005).Minimal recursion semantics: an introduction.Research on Language and Computation, 3(4):281–332.


introduction to grammar engineering with hpsg project ...yzhang/ge-ss08/lecture-01.pdf · lecture...

Documents