Another ACL'99 Workshop Announcement

[Priscilla Rasmussen]

 =========================================================================

				SIGLEX99
			Standardizing Lexical Resources
			    June 21, 22, 1999
			 University of Maryland

 ==========================================================================
       	   		FIRST CALL FOR PAPERS
 =========================================================================

As our national interests become increasingly global, timely access to
information becomes more and more necessary.  Many promising strategies for
information provision rely heavily on lexical resources, including ontologies.
Our next major challenge is providing a standardized lexical resource: an
inventory of word meanings, or senses, associated with criteria for
distinguishing them.  Currently there are several different on-line lexical
resources that are being used for English, WordNet, Longman's, the Oxford
English Dictionary, (OED), CIDE from Cambeidge University Press (CUP), Collins,
and Webster's, to name just a few, and they each use very different approaches
to making sense distinctions.  Various computational lexicons and related
resources such as ontologies are under development, including the European
PAROLE/SIMPLE lexicons, the Generative Lexicon, the SENSUS ontology,
Mikrokosmos, WordNet, Framenet, and the theory of Lexical Conceptual
Structures.  Each takes a very different approach and makes reference to
different underlying theories of semantics.  This divergence of resources has
motivated the efforts of the EAGLES Lexical Semantics Group, which is defining
a common format for lexical semantic representation for 12 languages.
http://www.ilc.pi.cnr.it/EAGLES96/rep2/

In a recent evaluation of word sense disambiguation systems, SIGLEX98-SENSEVAL,
(also supported by Euralex, Elsenet, ECRAN and SPARKLE)
"http://www.itri.brighton.ac.uk/events/senseval",
the training data and test data were prepared using a set of Oxford University
Press (OUP) senses.  This made it difficult to evaluate the performance of
pre-existing systems that had been built using other lexical resources.  A
mapping was made from the OUP senses to WordNet senses, so that WordNet systems
could be included, but this was somewhat problematic as there were far fewer
WordNet senses, and frequently no direct mapping was possible.  As do most
dictionaries, OUP and WordNet often make different decisions about how to
structure entries for the same words which are all equally valid, but simply
not compatible.  Therefore, it becomes especially difficult to include
pre-existing systems in the evaluation that rely on a pre-existing lexical
resource other than the one used as the Gold Standard.  The question that
arises here is the likelihood of making performance preserving mappings between
lexical resources.  Is it even possible to treat one lexical resource as a
standard that other resources can be mapped to? (This is true even when
focusing on just one language - the problem simply becomes more explosive when
additional languages become involved.)  All of the participants in
SIGLEX98-SENSEVAL agreed that they would prefer evaluations based on running
text rather than corpus instances, but this is only feasible if the Gold
Standard sense inventory being used for tagging can be appropriately mapped
onto several different lexical resources.

The purpose of SIGLEX99 is to directly address the issue of standardization of
lexical resources, and performance-preserving mappings between existing
resources.  As a spin-off from SENSEVAL, we are investigating mapping the OUP
SENSEVAL senses onto other lexical resources.  We will also be tagging running
text with these senses, and other senses, and will circulate this ahead of time
to workshop participants.  There will be several working sessions focussed
around the mappings between lexical resources and the tagged samples.

Languages other than English will also be considered, in connection with
ROMANSEVAL, the subset of SENSEVAL for Romance languages (but with no
restriction to that language family). We will study the relevance of
EuroWordNet (EWN) sense dictinctions for WSD systems, and the applicability of
the Interlingua Language Index (ILI) created within EWN for cross-language
sense-standardization. An issue of particular interest is the mapping of
existing resources to the ILI, which could be an important step towards the
development of a standardized multilingual lexicon for WSD. Such a multilingual
gold standard could in turn be used to semantically tag parallel texts and thus
create standardized corpora useful for many multilingual applications. There
will also be a session to discuss the future of American involvement in EAGLES,
and how the workshop results and conclusions can be incorporated.

We will have invited talks on ontologies and lexical resources, and we welcome
submissions on any areas in lexical semantics and computational lexical
semantics, but particularly on the acquisition and use of lexical resources and
ontologies and on word sense disambiguation.  There will be a workshop
proceedings, and as we have done with our last two workshops, we will encourage
partipants to make electronic versions of their papers available on the web
prior to the workshop.  Likely invited speakers include Patrick Hanks (Oxford
University Press), Chuck Fillmore (Berkeley), and someone speaking on WordNet
or EuroWordNet and on SIMPLE (the European project for building harmonized
semantic lexicons for 12 European languages).


The schedule for paper submissions (ACL format, 6 pages):

SUBMISSION DEADLINE:				March 29, 1999

NOTIFICATION OF ACCEPTANCE:			May 7, 1999

CAMERA READY COPIES (and copyrights) DUE:	May 28, 1999


Please send submissions, hard copy or electronic (.ps or .doc), to:

Martha Palmer
Institute for Research in Cognitive Science
400A, 3401 Walnut Street/6228
University of Pennsylvania
Philadlephia, PA 19104
Telephone: (215) 898-0361
FAX No.: (215) 573-9247
e-mail: mpalmer at cis.upenn.edu

Program Committee:

Nicoletta Calzolari, Istituto di Linguistica Computazionale, Pisa
Bonnie Dorr, University of Maryland
Chuck Fillmore, University of California, Berkeley
Ralph Grishman, New York University
Patrick Hanks, Oxford University Press
Eduard Hovy, USC Information Sciences Institute
Nancy Ide, Vassar College
Adam Kilgarriff, ITRI, University of Brighton
Marc Light, MITRE Corporation
Martha Palmer, University of Pennsylvania, CHAIR
James Pustejovsky, Brandeis University
Philip Resnik, University of Maryland
Patrick St Dizier, IRIT-CNRS, Université Paul Sabatier
Antonio Sanfilippo, European Commission, DG XIII
Frederique Segond, Xerox Research Centre, Grenoble
Jean Véronis,  Université de Provence
Evelyne Viegas, New Mexico State University
Piek Vossen, University of Amsterdam
Yorick Wilks, University of Sheffield
David Yarowsky, John's Hopkins University
Antonio Zompolli, Istituto di Linguistica Computazionale, Pisa