* * * C A L L F O R P A P E R S A N D C H A L L E N G E S * * *
PASCAL Workshop on Learning Methods for Text Understanding and Mining
January 26-29, 2004
Grenoble (France)
Important facts:
- Abstract of scientific contributions, submission due: November 28, 2003
- Challenge proposals, submission due: December 5, 2003
INTRODUCTION
------------
PASCAL (Pattern Analysis, Statistical Modelling and Computational Learning) is
the name of a Network of Excellence sponsored by the European Union as part of
its IST program. It brings together experts from basic research areas such as
Statistics, Optimisation and Computational Learning and from a number of
application areas, with the objective of integrating research agendas and
improving the state of the art in all concerned fields.
As part of its activities, the PASCAL network organises a workshop on the
subject of "Learning Methods for Text Understanding and Mining". The aim of the
workshop is twofold:
- Introducing to experts in statistics, computational learning and optimization
problems issuing from text understanding and mining which are both relevant and
suitable to be tackled within their framework;
- Proposing "challenges" (i.e.: concrete benchmark tasks) that will help
measuring improvements in the state of the art.
THE WORKSHOP
------------
In order to achieve these objectives, the Workshop will be organised as
follows:
- Jan 26 (afternoon only): Pre-workshop: Presentation of the results of the EU
IST project KerMIT ("Kernel Methods for Images and Text",
http://www.euro-kermit.org).
- Jan 27: Tutorials
- Machine Learning applied to Text Analysis: Overview (E. Gaussier)
- Memory-based Language Processing (W.Daelemans)
- Text Mining (D.Mladenic and M.Grobelnik)
- Kernel Methods for Natural Language Processing (J-M. Renders)
- Jan 28: Contributed scientific talks
- Jan 29: Challenge proposals and discussion
We anticipate that participants might attend only part of the workshop.
SCIENTIFIC CONTRIBUTIONS
------------------------
For the day of January 28, submissions of abstracts are invited in the
following areas of interest:
- Machine learning of phonology, morphology, syntax, semantics and translation
models
- Learning approaches to Document Retrieval, Categorization, Filtering and
Clustering
- Text mining
- Learning approaches leveraging document structure
- Machine Learning for Information Extraction
- Unsupervised and semi-supervised learning for Natural Language
Of special interest are contributions addressing linguistic components less
commonly made the object of Machine Learning approaches (e.g.: compositional
semantics), as well as contributions addressing the simultaneous learning of
multiple linguistic components.
In order to foster fruitful discussions and eventually collaborations between
the scientific communities represented at the workshop, scientific
contributions should, whenever possible, emphasize the limits of the approaches
described, and explicitely mention what difficult and important problems remain
to be solved, if any.
Selected presentations will be allocated slots of 30 minutes. Presentation
abstracts should be up to 4 pages long, in PDF or PS format, and suitable to be
printed on A4 paper. They should be sent by e-mail to Nicola Cancedda at the
address:
Nicola.Cancedda@xrce.xerox.com
CHALLENGES
----------
For the day of January 29, we invite submissions of proposals for PASCAL
challenges. The selected proposals will be presented in slots of 30 minutes
each in the morning, and will serve as a basis for the discussion that will be
held in the afternoon. Besides a description of the problem to be solved,
proposals should explicitely address:
- Format of the evaluation (TREC-like contrastive evaluation, permanent
web-based evaluation tool, ...);
- Public availability of data and other required resources;
- Estimated effort to build up resources, if any, not currently in the public
domain;
- Results already obtained on the data (if any);
- Key-words
We anticipate that some funding will be available from the PASCAL budget to
cover part of the expenses incurred in actually running challenges.
The PASCAL joint programme of activities also envisages the definition of
theoretical challenges. We thus also invite submissions of theoretical
questions and open problems relevant to the application of statistical learning
and optimisation to problems in Natural Language Processing, Information
Retrieval and Textual Information Access. Such proposals should provide,
besides the question itself, a justification of its relevance and a concise
overview of related available relevant results.
As for scientific contributions, proposals concerning tasks less commonly
addressed with Machine Learning techniques will receive special consideration.
Challenge proposals should be up to 4 pages long, in PDF or PS format, and
suitable to be printed on A4 paper. They should be sent by e-mail to Florence
d'Alché-Buc at the address:
IMPORTANT DATES
---------------
Please note the following deadlines:
- Abstracts of scientific presentations: November 28, 2003
- Challenge proposals: December 5, 2003
- Notification of acceptance: December 23, 2003
- Paper camera-ready deadline: January 16, 2004
- Workshop date: January 26-29, 2004
SPONSORSHIP
-----------
The workshop will be partly funded by a grant from the European Network of
Excellence "PASCAL".
ORGANIZERS
----------
* Nicola Cancedda (Xerox Research Centre Europe)
Nicola.Cancedda@xrce.xerox.com
* Florence d'Alché-Buc (LIP6, University of Paris 6)
florence.dalche@lip6.fr
PROGRAMME COMMITTE
------------------
* Nicola Cancedda (Xerox Research Centre Europe, Grenoble, France)
* Alexander Clark (ISSCO/ETI, University of Geneva, Switzerland)
* Florence d'Alché-Buc (LIP6, University of Paris 6, France)
* Walter Daelemans (University of Antwerp, Belgium)
* Ido Dagan (Bar Ilan University, Israel)
* Eric Gaussier (Xerox Research Centre Europe, Grenoble, France)
* Cyril Goutte (Xerox Research Centre Europe, Grenoble, France)
* Marko Grobelnik (Jozef Stefan Institute, Ljubljana, Slovenia)
* Dunja Mladenic (Jozef Stefan Institute, Ljubljana, Slovenia)
* Jean-Michel Renders (Xerox Research Centre Europe, Grenoble, France)
This archive was generated by hypermail 2b29 : Mon Nov 03 2003 - 14:46:02 MET