COOPERATIVE KNOWLEDGE ACQUISITION & KNOWLEDGE REFINEMENT SYSTEMS
Professor Derek Sleeman, Department of Computing Science, University of Aberdeen

Abstract: In this talk, I will give a general overview of activities in my group, and will then describe selected projects in some detail.

I'll argue that given the state-of-the-art in most complex real-world domains, that many Discovery- and Knowledge Acquisition- Systems need to be co-operative. In particular, I'll discuss systems where the machine's principal role is to systematically analyse considerable amounts of data, and where the expert is only required to provide a relatively small number of decisions or additional pieces of information. In particular, I'll illustrate this approach by discussing the REFINER system which is able to detect inconsistencies in a set of cases and suggest how these inconsistencies can be removed. Secondly, I'll discuss the TIGON system which is able to infer diagnostic rules for a machine (gas turbine) given vast amounts of data collected by the on-board computer and small amounts of high-level information provided by the domain expert.

I'll conclude by outlining how the range of tools/techniques which we have produced over the last few years, are being applied to tasks in clinical medicine.

This seminar was held at the Department of Computer Science, Royal Holloway, University of London on 19 February 1997

back