Despite its huge success, the World Wide Web is suffering from a major
limitation: processing information on the Web has been the privilege of agents that can understand full natural language and graphics, i.e.
humans only. The role of computers on the Web is so far limited to the transport and display of information, and they give us almost no support
in dealing with the content of the information. The aim of the "semantic web" is to make at least some of the meaning of information on the Web
accessible to machines, enabling them to give us much more support in finding, combining and deriving information than what is possible now.
This talk outlines the main ingredients that are necessary to realise this grand vision of a semantic web. I report on what has been achieved
so far, what is likely to be achieved in the near future, and what are the remaining hard challenges for researchers in artificial intelligence
and computer science alike.
Robust Knowledge Representation
14:00, Wednesday, 22 January 2003
The logical roots of knowledge representation (KR) have given the field much of its strength: well formalised foundations with well understood
properties. However, the logical roots are also responsible for some of the major weaknesses in KR. Deduction procedures are typically crisp
(answers are either yes or no, with no intermediate values), abrupt (answers are available only at the end of the procedure) and inefficient
(computation time cannot be traded against the quality of the answer). These unfortunate properties severely limit the applicability of KR, and
also do not make a plausible model of realistic reasoning strategies. This talk presents some of our results towards robust forms of KR which,
although still firmly rooted in logic, display anytime behaviour: algorithms can be interrupted at any time during their computation, and
will provide an approximation of the final answer. We give both theoretical characterisations of such algorithms, as well as practical
results obtained in diagnosis and classification.
For further information
contact either:
Haycock Lecture Theatre (Room CS/103)
Department of Computer Science
University of York
About the Speaker: After studying mathematics and computer science in
Amsterdam, Frank van Harmelen moved to the Department of AI in Edinburgh. There he worked on both Socrates, a logic-based toolkit for expert
systems, and on proof planning for inductive theorem proving. He was awarded a PhD in 1989 for his research on meta-level reasoning. He moved
back to Amsterdam in 1990 where he worked on the use of reflection in expert systems, and on the development of the (ML)^2 language for
formally specifying knowledge-based systems. In 1995 he joined the AI Department at the Vrije Universiteit Amsterdam, where he holds the chair
in Knowledge Representation and Reasoning. His current interests include approximate reasoning, the semantic web,
and medical protocols. He is also active in the World Wide Web consortium, contributing to the standardisation of OWL, the ontology representation
language for the semantic web. His research results have been reported in over 100 research papers and in three books (on meta-level inference,
on logic in knowledge-based systems and on the semantic web).
About the Lecture Series:
This lecture series is sponsored and organised by the Department of
Computer Science at the University of York and the School of
Computing at the University of Leeds. Its purpose is to
promote the strong research interests that both departments have in
knowledge representation and reasoning.
Further information can be found at http://www.cs.york.ac.uk/aig/seminars/dist.html.
Department of Computer Science
University of York
York YO10 5DD
phone: +44 (1904) 432745
email: frisch@cs.york.ac.uk
School of Computing
University of Leeds
Leeds LS2 9JT
phone: +44 (113) 233 1708
email: dianec@comp.leeds.ac.uk