Conf: INLG

[Philippe Blache]

From:  Svetlana Sheremetyeva <lana at crl.nmsu.edu




                           CALL FOR PARTICIPATION
                           Analysis for Generation

                                 a Workshop

                             in conjunction with

                   The International Conference on Natural
                  Language Generation INLG'2000 (June 13-14)


                      June 12, 2000, Mitzpe Ramon, Israel

1. The reasons why the workshop is of interest at this time.

The last decade has seen an explosion in the work done in the field of
NLG with the emphasis on the development of independent NLG
applications rather than generation modules of MT systems. While it
seems natural to consider problems of analysis and generation as two
sides of a coin in such NLP applications as MT, researchers working on
"pure" generation systems sometimes treat problems arising at every
stage of generation--content specification, sentence planning, and
surface realization--as independent. Time may be ripe for examining
the mutual utility of analysis and generation in greater detail. The
impetus is, as can be expected, the goal of minimizing system-building
efforts in language engineering.

 2.  A brief technical description of the issues the workshop will address.

 The workshop proposes to address:

*  issues connected to the needs of analysis in generation systems
   as well as

*  issues of interrelation between generation and analysis from the
standpoint
   of reusability and adaptability of analysis techniques and tools for
   generation.

*  issues of reusability of analysis knowledge and methodology of its
   acquisition for generation.

Analysis as part of generation. A wide range of complex problems which
are considered to be specific for generation--content specification,
planning, and grammaticalization--may lead one to believe that
generation is completely independent of analysis. Most generators
tacitly assume that an intermediate system module can use the output
of the preceding module as its input without any processing (that is,
analysis) of this output. This is not, however, the case in practice.

A modicum of analysis is, in fact, an essential part of every
generation system.  The input to generation systems such as raw data
in tables, lists, diagrams, elements of various databases or even text
snippets that are fed into the system directly by a user still must
first be somehow processed, that is, analyzed. The analysis is needed
to provide both the necessary content and, often, format for the
content specification (see, for example, Dale 1995, Robin 1994, Kukich
1988, McKeown et al. 1994, Bateman and Teich 1995). This problem
becomes especially important in those applications (including
multilingual ones) in which at least some input to generation is in
textual form (e.g., Sheremetyeva and Nirenburg 1996).  Many NLG
systems often use a variety of analysis techniques. The question
arises whether it is possible to develop criteria to better choose and
integrate analysis techniques which could be efficiently applied at
different stages of generation.

Reusability and adaptability of analysis techniques and tools for
generation. While it is not uncommon to believe that generation and
analysis are not reversible, a number of contributions over the years
have discussed reversibility of analysis and generation resources,
especially the grammars. Appelt 1987, Barnett and Mani 1990 and van
Noord 1993, among others, demonstrate how the use of reversible
grammars may lead to efficient and flexible natural language parsing
and generation systems. It is worth discussing constraints on
reversibility.

Reusability of analysis knowledge and methodology of its acquisition
for generation. Generation and analysis are closely related in that
both processes use many similar resources, and often it is less
expensive to reorganize an existing "analysis" resource (e.g., an
analysis lexicon) than to acquire one for generation from
scratch. (Allgayer et al. 1989, Viegas and Beale 1996, Sheremetyeva
and Nirenburg 1999a, 1999b). It is worth discussing how to establish
whether a resource built for analysis can be used for generation and
at what price.


 THE PROGRAM:
June,12, 00

13.00 - 13.15      S. Sheremetyeva. Introduction

13.15-13.50        H. Manuelian. Reusability of Mental Representation Theory
                   in Automatic Generation: Generating Expressions Referring
                   to Objects  with G-TAG and RMs

13.50-14.25        D. Lonsdale. Leveraging Analysis Operators in Incremental
                   Generation

14.25-15.00        S. Busemann. Interfacing Constraint-Based Grammars and
                   Generation Algorithms

15.00-15.20        Break

15.20-17.00        Panel :  R. Kempson, S. Nirenburg, M.Zock,
S.Sheremetyeva.


PLEASE FIND DETAILS ABOUT REGISTRATION AND ACCOMMODATION at the main
conference homepage at http://www.cs.bgu.il/~nlg2000


 Organizing Committee

 Svetlana Sheremetyeva, Chair and contact person
 Computing Research Laboratory, New Mexico State University, USA
 lana at crl.nmsu.edu

 Sergei Nirenburg
 Computing Research Laboratory, New Mexico State University, USA
 sergei at crl.nmsu.edu

 Richard Kittredge
 Department of Linguistics and Translation, University of Montreal
 kittredge at IRO.UMontreal.CA

 Anna Sagvall Hein
 Department of linguistics, Uppsala University
 Anna at ling.uu.se

 Evelyne Viegas
 Microsoft Corporation
 evelynev at microsoft.com

 Michael Zock
 Language & Cognition LIMSI - CNRS
 zock at limsi.fr

___________________________________________________________________
Message diffusé par la liste Langage Naturel <LN at cines.fr>
Informations, abonnement : http://www.biomath.jussieu.fr/LN/LN-F/
English version          : http://www.biomath.jussieu.fr/LN/LN/
Archives                 : http://web-lli.univ-paris13.fr/ln/