FUZZ-IEEE 2001Workshop on Modelling with Words University of Melbourne, Australia,  2nd/3rd December 2001

SCIENTIFIC DESCRIPTION

This workshop will provide a forum to share research on the emerging field of "modelling with words", a field at the intersection of fuzzy information granulation and machine learning. This field has built upon the computing with words paradigm originally introduced by Zadeh to capture the idea of computation based on linguistic terms rather than numerical quantities. However, where computing with words has focused on inference from linguistic knowledge bases, the focus of modelling with words has been in acquiring/learning such models. One of the distinguishing features of this new field is that it decomposes information spaces into imprecise regions or fuzzy granules that are subsequently used to model systems. These models may either be learnt from example data or provided by human experts or a combination of both. In fact, the fusion of these sources of information plays a central role in modelling with words. Typically, the acquired systems aggregate the granular information using probabilistic or fuzzy logic reasoning techniques.

Recent work has demonstrated (with approaches such as fuzzy decision trees, Cartesian granule feature modelling, weighted rules and fuzzy prototyping) that the modelling with words paradigm enhances both model tractability and transparency on the one hand and generalisation power on the other. To-date, however, this work has typically been limited to small world problems (with tens of features/attributes/variables). One of the key challenges that lies ahead for this paradigm is the issue of scalability to large problem domains (such as text categorisation). In scaling these approaches, how can transparency be maintained and possibly enhanced? The need to model larger scale and more complex systems in a transparent way necessitates the development of feature selection techniques as well as other methods of finding appropriate sub-models and then combining them. Other issues that need further research include which words can be used to partition information spaces? Are there limits on the granularity? How can granular models be merged? Can this paradigm accommodate incremental learning? What is the best formal framework for learning and representing linguistic models?

It is hoped that the submissions to this workshop will address these and other issues that provide not only a challenge for the paradigm modelling with words, but also an interesting future for this field. Both theoretical and applied contributions are welcome (examples of problem domains include information retrieval, computer vision, decision support systems, profiling etc.)
 
 
 

ORGANIZING AND PROGRAM COMMITTEE Jonathan LAWRY Advanced Computing Research Centre Department of Engineering Mathematics University of Bristol, BS8 1TR, UK E-mail: J.Lawry@bris.ac.uk Tel: +44-117-9288184 Fax: +44-117-9251154 James G. SHANAHAN Xerox Research Centre Europe (XRCE) Grenoble Laboratory  6 chemin de Maupteruis, 38240 Meylan, France e-mail: Shanahan@xrce.xerox.com WWW:http://www.xrce.xerox.com/people/shanahan/ Tel: +33-476-615113 Fax: +33-476-615099 IMPORTANT DATES Extended Abstract Submission: June 1, 2001 Notification of Acceptance: August 1, 2001 Final Paper Submission Due: Septemeber 1, 2001 Workshop Dates: December 2/3, 2001

SUBMISSION DETAILS

Papers must be submitted electronically in postscript format to the session organization (see addresses below). Please follow the guidelines for paper submission outlined by IEEE at http://www.ieee.org/. Final paper submissions must not exceed 6 printed pages, including all figures and tables.