Linguistic specifications for typed feature structure formalisms

Directorate-General For Translation European Commission

Description

Information technology and telecommunications
Information - Documentation

Sujets

• Information processing
• Linguistics
• Machine translation
• Programming language
• Terminology

Informations

• Publié par: EUROPEAN-COMMISSION
• Nombre de lectures: 8
• Langue: English
• Poids de l'ouvrage: 6 Mo

Extrait

:* HU n ι SK« IÍIIÚ r¿T« : ι ICI a H H ;f¿víci ir±v H MIC
AND NATURAL LANGUAGE PROCESSING
Volume 10
JNGUISTIC SPECIFICATIONS FO
PED FEATURE STRUCTURE FORMAL
Edited by
Frank VAN EYNDE and Paul SCHMIDT
European Commission
*** Studies in machine translation
and natural language processing
Published by:
Office for Official Publications
of the European Communities Managing Editor
Erwin Valentinì (European Commission), Luxembourg
Editorial board
Doug Arnold
(Department of Language and Linguistics, United Kingdom)
Nicoletta Calzolari
(Istituto di Linguistica Computazionale, Italy)
Frank Van Eynde
(National Fonds voor Wetenschappelijk Onderzoek, Belgium)
Steven Krauwer
(Rijhuniversiteit Utrecht, Netherlands)
Bente Maegaard
(Center for Sprogteknologi, Denmark)
Paul Schmidt
(Institutför Angewandte Informationsforschung, Germany)
Luxembourg: Office for Official Publications of the European Communities, 1998
ISSN 1017-6568
© European Communities, 1998
Printed in Italy Volume 10
Linguistic Specifications for
Typed Feature Structure Formalisms
Edited by
Frank Van Eynde and Paul Schmidt
European Commission Volume 10
Linguistic Specifications for
Typed Feature Structure Formalisms
Editors
Frank Van Eynde
Paul Schmidt
Contents
FRANK VAN EYNDE AND PAUL SCHMIDT
Introduction 7
PAUL SCHMIDT
Formal Assumptions 11
STELLA MARKANTONATOU AND LOUISA SADLER
Lexical Generalisations 39
PAUL BENNETT AND PAUL SCHMIDT
Phrase Structure 70 TONI BADIA AND CARME COLOMINAS
Predicate-Argument Structure 137
FRANK VAN EYNDE
Tense, Aspect & Negation 209
VALERIO ALLEGRANZA
Determination and Quantification 281
FIAMMETTA NAMER
Support Verb Constructions 315
References 333
Names and Adresses 344 Introduction
Frank Van Eynde and Paul Schmidt
This volume is an abridged and revised version of the final report of the project "Inves­
tigation of Linguistic Specifications for Future Industrial Standards". This project was
part of the Multilingual Action Plan of the European Union (1994-1995) and was carried
out by a consortium of seven research institutes, i.e. IAI, Saarbrücken; Gruppo DIMA,
Torino; University of Nancy; University of Essex, Colchester; Universität Pompeu Fabra,
Barcelona;y of Leuven; UMIST, Manchester.
The purpose of the project was the development of linguistic specifications for TFS-based
formalisms, i.e. formalisms which make use of typed feature structures (TFS) for the rep­
resentation of linguistic information. The results of this development are presented and
exemplified extensively in the final report (Schmidt, Bennett, Theofilidis, Markantona-
tou, Sadler, Badia, Colominas, Namer, Allegranza, Van Eynde, Schoenmakers and Tuells
(n.d.)). These specifications are not geared to any particular language, but provided a
starting point for the development of more detailed specifications for a number of individ­
ual languages, more specifically Danish (Braasch, Pederson, Joergensen, Povlsen, Paggio
and Underwood (1995)), Dutch (Schuurman and Schoenmakers (1995)), French (Heyd,
Jacquey, Namer and Recourcé (1995b)), German (Reuther, Rieder and Theofilidis (1995)),
Greek (Charalabopoulou, Efthemiou, Gianoutsou, Malagardi, Moschonas and Samiotou
(1995)) and Italian (Allegranza, Mazzini and Ruimy (1995)). These were the object of
separate monolingual projects and provided in turn the point of departure for a number
of language specific implementations, notably in the framework of the LS-GRAM project
(Large Scale Grammars for EU languages, Linguistic Research and Engineering). The
LS-GRAM modules, which also include English, Portuguese and Spanish grammars, are
written in the ALEP formalism, but the linguistic specifications which will be presented in
this volume are not tuned to the specific properties and particularities of this formalism;
instead they aim at compatibility with any kind of TFS-based formalism.
Apart from compatibility with this family of NLP formalisms there are a number of further
requirements which the specifications aim to fulfill, such as breadth of coverage, internal
coherence and multilingual orientation. On a more technical level the specifications are
required to be monostratal, lexicalist and constraint-based. Since these are the main
characteristics of Head-driven Phrase Structure Grammar, it will not come as a surprise that Studies in MT and N LP, Volume 10
HPSG forms the point of departure for most of the work which will be presented in this
volume. At the same time, it should be added that the resulting specifications do not all fit
in the mould of Pollard and Sag (1987) and Pollard and Sag (1994). This is mainly due to
two reasons.
First, in order to enhance the usability of the specifications for processing purposes we have
limited the use of some of the HPSG tools which are less attractive from a computational
perspective, such as relational dependencies and lexical rules. The effort to reduce the use
of such computationally expensive tools is characterized in this volume as 'aiming for a
lean approach'. It is explicitly addressed in the chapters on formal assumptions, lexical
generalisations and phrase structure, and it is implicitly adopted in the other chapters.
Second, in order to attain a reasonably wide coverage we have to deal with phenomena
which have so far received little attention in current HPSG, such as the semantics of tense
and aspect, the treatment of support verbs, the semantics of bare plurals and the predicate-
argument structure of nonverbal elements. The proposals which are made for these topics
are hence, by necessity, novel.
In order to give a more precise idea of what the volume contains we now provide a brief
survey of the different chapters.1
1. Formal Assumptions: This chapter describes the 'lean HPSG' version which will
be assumed throughout the volume. Given the general aim of developing linguistic
specifications for TFS-based formalisms it is only natural that most of the chapter
deals with the formal properties of typed feature systems. The main topics of attention
concern
• Structured types vs. types defined by appropriacy conditions
• Type inheritance: single vs. multiple inheritance, monotonic vs. default inher­
itance
t Architectural properties like the Bounded Complete Partial Ordering (BCPO)
restriction
Other topics which will be dealt with concern the use of logical connectors in feature
structures and the use of macros as a means for simulating structured types.
2. Lexical Generalisations: This chapter discusses the various techniques and tools
which have been proposed in the literature to express lexical generalisations in a
typed monotonic inheritance framework. These include multiple monotonic inheri­
tance, lexical rules, underspecification, macros and distributive disjunction. For each
technique and/or tool, the following issues are explored:
• The linguistic intuitions that each technique encodes
• The expressive power of the system that is necessary to support each technique
'The order of the chapters is not arbitrary, but each chapter is self-contained and can be read in isolation.