Wednesday 20th October 1999 (6 pm)
ABSTRACT
The acquisition of Knowledge Bases (KBs) for use with Expert Systems/Decision Support Systems has long been
recognised as a difficult and time-consuming task. More recently it has been appreciated that if a KB is complex
it is likely to be both incomplete (& so more Knowledge needs to be acquired) & inconsistent (when the
KB needs to be refined). Additionally, for complex domains it is unlikely that a complete domain theory will be
available, hence the need to involve the domain expert as an oracle & the need for Co-operative Knowledge Acquisition
& Refinement Systems. Several examples of such systems which use a variety of knowledge representational schema
& which have been cited in a number of domains (e.g. selection of wines, diagnosis of turbines, stock control
& botanical taxonomies etc) will be reviewed.
The focus of the talk is a current project to build a (Co-operative) Workbench so that clinicians can predict outcome
for patients given real-time monitoring data & (incomplete & inconsistent) background Medical Knowledge.
Specifically we are working with data-sets for serious head injury patients which have been collected in the Intensive
Care Unit at Western General Hospital in Edinburgh. The overall project has been conceived as 3 phases:
Finally, this work will be related to current research themes in Knowledge Technology - such as the design & refinement of Problem Solving Methods, Reuse of Knowledge Bases, and the semi-automatic acquisition of KBs from texts.
DEREK SLEEMAN trained as a Physical Scientist at King's College London, and moved to Leeds where he was a Lecturer
in Computing & co-founded the Computer-Based Learning Unit in 1969. This lead to his interest in Intelligent
Tutoring Systems & an edited volume on that subject with John Seely Brown in 1981. He moved to Stanford in
1982 where he was a Senior Research Associate at the Heuristic Programming project & Associate Professor of
AI & Education.
He returned to Aberdeen in 1986 where he was appointed the University's first Professor of Computing Science. His
Research activities have remained at the intersection of AI & Cognitive Science, but the focus has moved from
ITSs to Co-operative Knowledge Acquisition & Knowledge Refinement Systems. For 3 years he was the Academic
Co-ordinator of the EU's Network of Excellence in Machine Learning & Knowledge Acquisition; additionally he
has been a Program Committee member for the International/European/National Conferences in Machine Learning &
Knowledge Acquisition. He has also served on various Editorial boards including the Machine Learning Journal &
the International Journal of Human-Computer Studies.
The Evening Lectures are free to both members and non-members of SGES.
For further information contact:
Dr. Hui Liu, Department of Computer Science, Birkbeck College hui@dcs.bbk.ac.uk