Newsletter 1/2002
From SIGSEM
----------------------------------------------------------------------
SIGSEM Newsletter 1/2002
----------------------------------------------------------------------
ACL Special Interest Group in Computational Semantics (SIGSEM)
SIGSEM Webpage: http://www.sigsem.org
Current SIGSEM Membership: 542
----------------------------------------------------------------------
CONTENTS
1. SCANALU 2002
2. Computational Semantics Website
3. Report on ICoS-3
4. ICoS-4
5. IWCS-5
6. Information on Computational Semantics Projects
7. Working Group on Multimodal Meaning Representation
8. Report on Helsinki Computational Semantics Course
9. Next Newsletter
----------------------------------------------------------------------
1. SCANALU 2002
SIGSEM is endorsing the First International Workshop on Scalable
Natural Language Understanding SCANALU 2002, which will be held on May
23-24 2002, Villa Bosch, Heidelberg, Germany. The workshop webpage is
at http://www.eml.org/SCANALU, and can also be accessed from the
SIGSEM homepage (www.sigsem.org).
----------------------------------------------------------------------
2. COMPUTATIONAL SEMANTICS WEBSITE
Kyle Rawlins has up a website about computational semantics at
http://mas.cs.umass.edu/~rawlins/clsbib/
which may be of interest to SIGSEM members.
----------------------------------------------------------------------
3. REPORT ON ICOS-3
ICoS-3, the third international workshop on computational semantics
was held from the 18th to the 19th of June 2001 in Siena,
Italy. Unlike previous ICoS workshops it was not a stand-alone event:
it was a satellite workshop of IJCAR, the International Joint
Conference in Automated Reasoning. You can find a report on ICoS-3,
written by Raffaella Bernardi, on the SIGSEM homepage
(www.sigsem.org).
----------------------------------------------------------------------
4. ICOS-4
ICoS-4, the next workshop in the inference in computational semantics
series, will be held sometime in 2003. A number of proposals are
currently being discussed, including holding it as a satellite
workshop of an ACL event. When known, details will be posted on the
SIGSEM homepage (www.sigsem.org) and in this newsletter. In the
meantime, there's IWCS-5 to look forward to....
----------------------------------------------------------------------
5. IWCS-5
The Fifth International Workshop on Computational Semantics (IWCS-5)
will be held on January 15-17, 2003, in Tilburg, The Netherlands. It
will be hosted by the Computational Linguistics and Artificial
Intelligence department at Tilburg University. The event is endorsed
by SIGSEM. Following the tradition established by IWCS-1 through
IWCS-4 (see http://pi0239.kub.nl/~sigsem/iwcs4.html), the workshop
aims to bring together researchers interested in any aspects of the
computation of meaning in natural language or in language-based
multimedia objects. A Call for Papers will soon be distributed, with
a submission deadline in September 2002.
----------------------------------------------------------------------
6. INFORMATION ON COMPUTATIONAL SEMANTICS PROJECTS
At the last SIGSEM business meeting (during IWCS in Tilburg, January
2001), we agreed the SIGSEM website could and should do more for
SIGSEM members than merely to list them. One of the things we agreed
would be useful are links to projects that have something to do with
computational semantics. For example, projects addressing any
semantic/pragmatic issue in discourse and dialogue processing, and so
on.
Now, help from the members is needed to achieve this! Please send to
info@sigsem.org information about computational semantics projects you
are involved in or know about, in the following format:
Name of project (plus acronym if any): <name, acronym>
Project's web address: <url>
Project's life-span: <year1-year2 or ongoing>
The project's aim/scope (max. 50 words): <description>
Not only links to currently running projects are useful, but also to
completed projects that produced results which remain interesting for
the community. For example, Trindi, Verbmobil, DYANA or FraCAS are
probably such past projects that remain relevant.
Looking forward to receive your input!
Ivana Kruijff-Korbayova
----------------------------------------------------------------------
7. WORKING GROUP ON MULTIMODAL MEANING REPRESENTATION
>From 29 October - 2 November a very interesting workshop took place at
the Dagstuhl conference center in Germany on the topic of "Fusion and
Coordination in Multimodal Interaction". The organizers had invited
some 30 experts in various aspects of multimodal dialogue, with an
emphasis on speech, gestures and facial expressions, who discussed
recent developments in this area, and participated in four working
groups concerned with the following topics:
- data collection and multimodal annotation tools;
- multimodal meaning representation;
- software architectures for multimodal dialogue systems;
- research agenda for multimodality.
The working group on multimodal meaning representation consisted of
John Lee (Edinburgh), Fabio Pianesi (Trento), Noelle Carbonell
(Nancy), Paul Mc Kevitt (Londonderry), Catherine Pelachaud (Rome),
Laurent Romary (Nancy) and Harry Bunt (Tilburg, chair), and benefited
from additional contributions by Justine Cassell (Cambridge, MA),
Emiel Krahmer (Tilburg) and Wolfgang Wahlster (Saarbruecken). This WG
came to the conclusion that it would make sense to try to develop
recommendations for how to represent the meanings of communicative
actions involving speech and/or gestures and/or facial expressions,
and possibly other forms of nonverbal communication, in a way that
allows the construction of integrated ("fused") multimodal meaning
representations, while doing justice to the specific properties of
each of the modalities involved. Some initial constraints and
requirements for such representations were identified, as well as ways
to delimit the task of further developing recommendations. The notes
of the WG, made by Laurent Romary, are available at
http://www.dfki.de/~wahlster/Dagstuhl_Multi_Modality/
It was felt in particular that this task could well be taken up
further in the form of a SIGSEM working group, since meaning
representation (especially in dialogue systems) very much goes to the
heart of computational semantics. Such a working group should be open
both to SIGSEM members, benefiting from the expertise of computational
semanticists, and to researchers in nonverbal dialogue who may not
(yet) be SIGSEM members. Laurent Romary and I would be willing to draw
this cart. Operating under the auspices of SIGSEM would make it
possible to use the SIGSEM web pages and newsletter for communication
about the activities of the WG, and to use SIGSEM-sponsored events
like IWCS and ICoS as opportunities for the WG to meet and to present
their activities to the computational semantics community. This WG
could also pave the way for possible recommendations to be then taken
up by the new committee at ISO (TC37/SC4) intending to provide basic
standards in the domain of language resources. We discussed this idea
with SIGSEM chair Patrick Blackburn, who fully supports this, as do
the other SIGSEM officers.
Anyone who is interested in actively contributing to the activities of
this working group, or has specific suggestions for how to organize
its activities, please contact us at harry.bunt@kub.nl.
Harry Bunt and Laurent Romary
----------------------------------------------------------------------
8. REPORT ON HELSINKI COMPUTATIONAL SEMANTICS COURSE
The course `An Introduction to Computational Semantics', lectured by
Patrick Blackburn and Johan Bos, was held at the 13th European Summer
School in Logic, Language and Information (ESSLLI), Helsinki, 13-24
August, 2001 (see http://www.coli.uni-sb.de/~bos/comsem/esslli.html).
The course was sponsored by the EACL, who covered Johan Bos's travel
and accommodation costs.
The course was a foundational course. That is, it was designed to
cover the basics of the subject in a manner comprehensible to students
from diverse disciplines. For this reason, the course was one of the
few two-week courses offered at the Helsinki ESSLLI: 10 sessions of 90
minutes each were held. Teaching computational semantics involves
introducing the student to a wide range of topics, explaining how they
all fit together, and showing how to turn theory into computational
reality. This is not something that can be rushed.
The course focussed on two themes: representation and inference. On
the representation side, we covered basic ideas and techniques
(compositionality, lambda-driven semantic construction, treatment of
scope ambiguities), initially in the setting of first-order logic, and
then in the setting of Discourse Representation Theory (DRT). We
discussed both the underlying theory and showed how to implement
semantic construction tools in Prolog. On the inference side, we
introduced the basic ideas of logical deduction and related them to
concrete linguistic problems. At the end of the first week we brought
the themes of representation and inference together for the first
time: we showed how to construct a system which which drew on freely
available inference tools (such as theorem provers and model-builders)
to construct first-order semantic representations for sentences and
short texts. At the end of the second week we brought them together
in a deeper way: we showed how to use such tools to implement van der
Sandt's DRT-based presupposition algorithm.
By and large, the course seems to have been a success. Attendance
peaked at 120, and 90 stayed on till the end. The sessions were a
blend of straight lectures, system demos, and discussing the answers
to assignments. Students from a wide variety of backgrounds attended
(even computational biology was represented) and feedback was
generally positive.
We would like to thank the EACL for their financial support (it made a
big difference) and our students for their helpful feedback on the
material presented.
Patrick Blackburn and Johan Bos
----------------------------------------------------------------------
9. NEXT NEWSLETTER
The SIGSEM newsletter is sent only to SIGSEM members, and by default
it is sent to all members. If you don't want to receive it again,
please send an email saying so to:
info@sigsem.org
We anticipate sending out the newsletter 3 times a year, at 4 month
intervals.
----------------------------------------------------------------------
SIGSEM Newsletter 2/2001 was edited by:
Patrick Blackburn, INRIA Lorraine.
(patrick@aplog.org, http://www.aplog.org)
----------------------------------------------------------------------
>> back to Newsletters
