Goal
Topics
Papers
Topics
Dates
Speakers
Registration and practical issues
Committees
General goal of the workshop
Research in grammatical inference is presented in the main conference of the field (the International Colloquium in Grammatical Inference) which is held every 2 years, but also in a variety of conferences covering a range of disciplines: machine learning, computational linguistics, web engineering, bio informatics etc.
The goal of the workshop is to provide a meeting point where researchers from the field or interested in the field can discuss some of the topics that have appeared to be of special importance. It is hoped that the right combination of tutorial-like presentations of established work, of recently published material and of ongoing research will provide a forum for exciting ideas.
The four topics that have been chosen for discussion are those of Learning in a noisy setting, learning classes above the context free learning distributions over strings and transducers. As these themes only cover part of the field, a special open session will take place.
In order to attract a large audience, and specially that of students and young researchers, the costs have been kept low.
Back to Top
Topics
Grammatical inference researchers have mostly focused their attention on a number of traditional tasks. In order to be able to tackle some of the challenges we should face a certain number of questions want specific
attention: stochastic languages, transducers, beyond context-free grammars or dealing with noise, for instance.
The goal of the workshop is to get an updated picture of some of these key issues:
- Learning with noise: noisy labels, corrupted strings, learning a noise model, learning with noisy queries. What are the algorithms, the specific convergence results?
- Upward from context free. There are a number of ways of considering languages that are capable of describing patterns and rules that are not context-free... String kernels, linear and semi-linear sets, rewriting systems, mildly context-sensitive languages.
- Stochastic languages: there are now a number of algorithms that learn stochastic finite state automata, deterministic or not, but the picture is still unclear: which would work better, for small or
large alphabets?, in an approximation or an identification task?
- Transducers. From the very theoretical results of the nineties to the more experimental approaches
of the Tenjinno competition, it would be interesting to study if and how the methods are applicable,
for instance for morphology learning tasks.
The workshop will be organised into 5 half-day sessions, 4 of which will cover the above themes. A fifth
session will be left free in order to discuss alternative topics. In each session there will be a series of
presentations and a panel.
Back to Top
Selection of papers
The workshop is intended to be a meeting point between researchers in the field over a selected number of items. We therefore encourage submissions of abstracts covering:
- Previously published material, specially when this has been presented to conferences outside the normal ICGI cycle.
- Preliminary work, specially by PhD students. This can be the occasion to present ongoing work to specialists of the field and receive immediate feedback!
- New work not yet published. The workshop can perhaps help strengthen the results...
If in doubt about the fact that your work is suitable for presentation please contact the organiser or the area chairs.
Submission:
Please send a two-page abstract (including figures and references) to cdlh AT univ-st-etienne.fr
All papers must conform to the ICML06 style
Presentations will be selected depending on relevance to the workshop topic, quality, and novelty.
Selected papers of the workshop will be considered for publication in a special issue of an international journal.
Back to Top
Dates
Deadline for abstract submission: October 15, 2006
Notification of acceptance: October 25, 2006
Deadline for early registration: October 31, 2006
Conference dates: November 20-22, 2006
The above dates are
firm.
Back to Top
Invited Speakers
Researchers that have confirmed their presence at the workshop:
- Alex Clark, University of London, UK
- Jose Oncina, University of Alicante, Spain
- Takashi Yokomori, Waseda University, Japan
Back to Top
Committees
Area chairs
- Probabilistic languages: Jean Christophe Janodet and Amaury Habrard
- Upward from Context-free: Leo Becerra Bonache and Colin de la Higuera
- Transducers: Alexander Clark and Rémi Eyraud
- Noise: Jose Oncina and Frédéric Tantini
- Special open session : Colin de la Higuera
Organising Committee
- Colin de la Higuera, University of Saint-Étienne
- Leo Becerra-Bonache, University of Tarragona
- Marc Bernard, University of Saint-Étienne
- Rémi Eyraud, University of Saint-Étienne
- Baptiste Jeudy, University of Saint-Étienne
- Jean-Christophe Janodet, University of Saint-Étienne
- Thierry Murgue, University of Saint-Étienne
- Frédéric Tantini, University of Saint-Étienne
Back to Top
Registration and Practical issues
Registration
The registration fee is:
- FREE for early registration (October 31, 2006)
- 50 EUR from November 1st to November 19th
- 100 EUR on-site registration
Registration, including the early one, must be accepted by the organisers, i.e. must come from a scientist working in the field or showing interest in the field.
The goal of the workshop is to allow participation by all those interested. In order to do this the organisers are proposing a free registration until october 31st. This is possible thanks to the help of the PASCAL Network. But it also means that the organisers may have to limit the number of registrations.
Registration should be done online at this address
Practical issues
The workshop will be located in the Laboratoire Hubert Curien, a new lab specialising in Optics, image processing and machine learning in Saint-Étienne.
You can get to Saint-Étienne by plane (to Lyon Saint-Exupery airport), by train or by car. Indications as how to get to the lab can be found on
this webpage.
Temperature in November is usually wet and fresh.
Sponsors
The workshop is sponsored by the EU's 6th Framework Programme
PASCAL Network of Excellence
, and specifically by the SIG UMCHI, and the International Community in Grammatical Inference.