Research Seminar - November 19, 2004

Dynamic, Ontology based Text Classification

Time: 11am, Friday 19th November

Computer Science & Software Engineering

Seminar Room 1.24

Speaker:
Dr. Albert Weichselbraun
Visiting Research Fellow

Abstract:

The rapid growth of online resources and the resulting information flood increasingly complicate the retrieval of information. This development has motivated the World Wide Web Consortium to reinforce their Semantic Web activities, with the goal to enrich current Web resources with meta data, that is readable and understandable for computers and agents. This meta data will be the foundation of the next generation Internet - the so called Semantic Web. Machine processable semantic information will facilitate more specific searches, ubiquitous computing, etc. as document processing will be based on an actual understanding of the content instead of simple keyword searches.

Semantic technologies have matured and important basic technologies like the final Web Ontology Language Recommendation have been specified. Although these technologies are supported by well developed libraries and toolkits, the usage of semantic data is still at a surprisingly low level.

The Semantic Web, like the Internet as a whole, unfolds its full potential through network effects - i.e. more semantic annotations lead to more tools supporting and evaluating those annotations, which increases the utility of semantic tags and therefore promotes the usage of more semantic data. Therefore one of the major challenges for the Semantic Web will concern the availability of semantic content - but convincing millions of users to annotate documents for the Semantic Web seems to be almost impossible. Automated classification attempts to overcome some of these problems.

The presentation will include a short introduction to the vision of the Semantic Web and the issues this extension of the World Wide Web will address. Technologies used for Ontology Based Text classification will be introduced followed by a discussion of advantages, shortcomings and planned extensions of the methods. The technical realisation of an automated classifier, including its integration within a webserver suite will be presented and finally the classifiers performance and applications using the deduced classifications will be discussed.

About the speaker:

Dr. Albert Weichselbraun is currently a visiting Research Fellow at the University of Western Australia jointly supported by the Business School and School of Computer Science and Software Engineering. After completing two Master degrees in Economics and Chemical Engineering, his doctoral thesis at the Vienna University of Economics and Business Administration focused on ontology-based text classification. In June 2004 he joined the webLyzard/EcoMonitor project which aims in the technical development of a platform to comprise and visualise spatial and local content propagation effects in Web data. His primary research interests are mathematical methods for text classification and identification, Semantic Web Technologies and Spam-defence techniques.