The Eurotra formal specifications

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: 14
• Langue: English
• Poids de l'ouvrage: 2 Mo

Extrait

STUDIES ΙΓΪ MACHINE TRANSLATION
AND NATURAL LANGUAGE PROCESSING
Volume 2
THE EUROTRA
FORMAL SPECIFICATIONS
Edited by
C. COPELAND, J. DURAND
S. KRAUWER, Β. MAEGAARD
* Commission of the European Communities Studies in Machine Translation
and Natural Language Processing
Published by:
Office for Official Publications
of the European Communities Studies in Machine Translation
and Natural Language Processing
MANAGING EDITOR
Erwin Valentim (CEC, Luxembourg)
EDITORIAL BOARD
Jacques Durand (University ofSalford, United Kingdom)
Frank van Eynde (Nationaal Fonds voor Wetenschappelijk Onderzoek,
België)
Tom C. Gerhardt (CRP-CU/CRETA, Luxembourg)
Steven Krauwer (Rijksuniversiteit Utrecht, Nederland)
Bente Maegaard (Center for Sprogteknologi, Danmark)
Karsten Strørup (CEC, Luxembourg)
Luxembourg: Office for Official Publications of lhe European Communities, 1991
ISSN 1017-6586
Catalogue number: CD-AM-91-002-EN-C
© ECSC-EEC-EAEC, Brussels · Luxembourg, 1991
Printed in the FR of Germany PREFACE
For a number of years, the Eurotra Machine Translation project has been mainly
known by a series of papers presented at conferences, short articles in journals or
books, small-scale demonstrations and internal reports leaked out to interested parties.
While in recent years the number of outside publications has risen sharply, as attested
in the bibliographies of the contributions to these volumes, it has been practically im­
possible for external readers and assessors to gain an accurate view of the Research
and Development work done within the confines of the project. The present series
Studies in Machine Translation and Natural Language Processing - published under
the auspices of the Commission of the European Communities - is intended to reflect
some of the research results which have been achieved in the last few years within Eu­
rotra, which officially ended in December 1990. The series will also cater for work be­
ing done or being planned in the follow-up programmes to Eurotra sponsored by the
Commission in partnership with the EC member states. The formal, linguistic and
computational specifications of the Eurotra system have been defined within succes­
sive versions of the Eurotra Reference Manual and reflected in implementations done
by language groups and official demonstration packages. But the Reference Manual
is not devised for external consumption: it has been written against a background of
assumptions shared by Eurotra co-workers and within a terminology apt to baffle the
uninitiated. Internal reports usually suffer from the same inevitable parochialism. As
a result, a great deal of the work done by all researchers in the Eurotra project runs the
risk of oblivion.
We hope that the present series will fill in this gap and offer an all-round view of the
research being carried out within the project and its follow-up.
The first two volumes, which must be read in close conjunction, are conceived as
background publications which give a description of the 'official' Eurotra system.
What is presented here is the basis of the development work and applied research done
by language groups roughly during the period 1988-1990. In Volume 1, entitled The
Eurotra Linguistic Specifications, the reader will first of all find a general introduction
to the Eurotra Programme by Bente Maegaard and Sergei Perschke. This first chapter
aims at giving a brief overview of the main goals of the Eurotra project and outlines
some of the salient characteristics of the Eurotra system. This introductory chapter is
followed by a fairly detailed article on thea Linguistic Specifications by Valerio
Allegranza, Paul Bennett, Jacques Durand, Frank van Eynde, Lee Humphreys, Paul
Schmidt and Erich Steiner. Two further articles complete Volume 1 - viz. a general
presentation of the Structure of Dictionaries in Eurotra by Pius ten Hacken, Bente
Maegaard and Heinz-Dieter Maas and a brief overview of Terminology in the project
by Jennifer Pearson and Dorothy Kenny.
Volume 2 offers a presentation of the Eurotra formal specifications. The first con­
tribution is a presentation of the Eurotra Framework by Annelise Bech. The following
article by Giovanni Malnati and Patrizia Paggio surveys the main features of the user
language. In the third and final contribution making up this second volume, Roberto
Cencioni offers a broad overview of the Eurotra software environment. It should be made abundantly clear that Volume 2, in particular, is not necessarily intended to be
read from cover to cover by the reader. Apart from their intrinsic interest, the articles
in Volume 2 provide a formal notation within which most other articles in Volume 1
and in a number of forthcoming volumes are couched. Piecemeal introduction of frag­
ments of notation by each contributor would have led to much duplication and not al­
lowed the reader to have an overall view of the formal machinery used in the Eurotra
project. We therefore hope that readers will be willing to familiarize themselves with
enough of this material to feel reasonably at ease with Volume 1 as well as forthcom­
ing volumes in the series.
The editors of these first two volumes are Charles Copeland, Jacques Durand, Ste­
ven Krauwer and Bente Maegaard.
A number of volumes are being planned on themes as varied as the assessment of
computational linguistic formalisms, interlevel processing, morphology, support
verbs and noun argument structure. The editorial board hopes that Studies in Machine
Translation and Natural Language Processing will constitute a bridge between the
Eurotra programme and the computational linguistics community. Vol.2
THE EUROTRA
FORMAL SPECIFICATIONS
EDITORS
Charles Copeland
Jacques Durand
Steven Krauwer
Bente Maegaard
CONTENTS
ANNELISE BECH
Description of the Eurotra Framework 7
GIOVANNI MALNATI, PATRIZIA PAGGIO
The Eurotra User Language 41
ROBERTO CENCIONI
The Eurotra Software Environment - A Broad Overview 113 ANNELISE BECH
Description of the Eurotra-Framework
Introduction
This paper describes the Eurotra-framework (called Ε-framework), the virtual ma­
chine underlying the Eurotra prototype software ETS. The purpose of the paper, the
first version of which originally appeared in the Eurotra Reference Manual 6.1 (inter­
nal document), is to give the reader a fairly detailed and somewhat formal introduction
to the notation in which current Eurotra grammars are conceived and implemented.
Before going into details about the Ε-framework, however, we will briefly present
some of the main factors which have determined the way it currently looks.
1. Historical Background
The Ε-framework has evolved over a couple of years, both in the sense that it is a
descendant of previous Eurotra formalisms, and in the sense that we have been fine-
tuning the core of the virtual machine and experimenting with improvements and spe­
cial purpose extensions to it.
As official developmental software, the predecessor of the Ε-framework was the
<C,A>,T Framework (Arnold et al. 1986, Arnold & des Tombe 1987). As the current
formalism is related to the <C,A>,T Framework, the reader familiar with the latter will
undoubtedly recognize similarities between the two frameworks as well as a number
of differences. We will not discuss the details of these aspects here (see e.g. Bech and
Nygaard 1988, and Bech et al. 1991), but constrain ourselves to a few general com­
ments on design issues.
Like its predecessor, the Ε-framework deliberately does not pledge strong allegiance
to one single existing linguistic theory, basically because of the well-known argu­
ments against importing a linguistic theory wholesale into an NLP system, (cf. e.g. the
discussion in: Shieber 1987), but also because of the external requirements to be sat­
isfied. From the linguistic point of view, the Eurotra formalism had to be an adequate
and reasonably intuitive means for the specification of Eurotra grammars for languag­
es from three different families. Even though several current linguistic theories make
claims of (near) universality, it is clearly the case that there are still many outstanding
issues; and perhaps more importantly - not until recently have the theories begun to be
applied extensively to a significant number of different natural languages.
Consequently, rather than incorporating the more debatable and controversial
linguistic principles into the definition of the virtual machine, we have generally opted 8 Studies in MT and NLP, Volume 2
for user-controllable solutions and for freedom in the choice of expression to permit
experimentation. The machine translation scenario in which the Ε-framework was to
be applied imposed the requirements that the formalism be suited for our
stratificational approach to fully-automatic translation of texts, and that it support a
high degree of modularity in the compositional view of the translation problem.
2. The Virtual Machine and the Eurotra Translation Theory
The Ε-framework is a unification-based special purpose tool developed for machine
translation within Eurotra. As s