Showing papers by "Nigel Shadbolt published in 1990"

Knowledge based knowledge acquisition

Abstract: A framework for knowledge based knowledge acquisition is presented. The ACKnowledge project addresses the problem of supporting the knowledge engineer in the phases of KBS development involving knowledge acquisition. The goal of ACKnowledge is to provide an integrated and principled system or knowledge engineering workbench (KEW). Within this project the author has been working on the problem of providing a coherent framework for the integration of knowledge acquisition (KA) methods and techniques. He has taken the view that KEW will act as a KBS for KA; this view has a number of consequences for the development of the next generation of knowledge acquisition tools.

Speaking About Plans

Abstract: An ESRC funded project1 has been looking at the interpretation and generation of cooperative dialogue between agents. The idea is to derive discourse from an underlying planning system — the thesis is that the fundamental driving force behind dialogue is the problem solving ability of the agent. The research has shown, via the construction of a number of computational models, that flexible dialogue can be obtained under this organisation. A number of important insights have arisen from the research which have led us to consider extensions to this work.

Hypermedia for knowledge engineering

Abstract: Discusses the implications for knowledge engineering of hypermedia systems. The author particularly focuses on the use of hypermedia for knowledge acquisition (KA) in all phases of KBS construction and maintenance. The acquisition of knowledge remains the critical phase of expert system development. A number of software support tools have appeared that provide for the analysis, refinement and integration of knowledge from diverse sources; texts, manuals, verbal transcripts, cases, and the experts themselves. Consideration of these systems reveals three classes of support tools. One class provides support for acquisition in specific domains. A second consists of computer implementations of particular knowledge acquisition (KA) techniques and are as such domain independent, the third class are loosely integrated systems. He also discusses how hypermedia technology is being used to facilitate various of these knowledge acquisition support tools.

Task driven specification of a knowledge acquisition tool for medical diagnosis (KAT-MED)

Abstract: Knowledge acquisition methodology is moving away, from reliance on the knowledge engineer, as a skilled interviewer, to the development of task and model driven tools, such as Mole [3], [5], Opal [6], More [5] and Knack (31. Such tools, based on strong domain models, are very useful in well structured areas such as medicine [6]. Our project investigates a more general approach based on the decomposition of expertise into generic tasks as advanced by [2], and the epistemological (four layer) description of conceptual/interpretation models of expertise in the KADS methodology [ l ] .

Process control supervision using qualitative models

Abstract: In modern process and power plant the cost of installation demands operation at peak efficiency for prolonged periods. This places significant demands on the control and monitoring systems to keep efficiency high while giving significant warning of a drop in efficiency or a component failure. We present an approach to process monitoring based on Qualitative Models which are used as a framework in which a range of monitoring techniques are located. The methods are described in the context of diagnosing faults in a heat exchanger. We present an application based on a Cogeneration scheme which highlights many of the issues in process monitoring.

Knowledge acquisition for knowledge based process control

Abstract: In previous papers (Shadbolt, Robinson and Stobart 1989a, 1989b) we have described the way in which process control can benefit from the techniques and methods of knowledge based systems. In Shadbolt et al 1989a we showed how knowledge intensive methods can be incorporated in supervising and automating the control engineering procedures associated with designing, commissioning and running closed loop control. Some of the most influential current approaches to knowledge acquisition conceive of the process as a modelling activity (Morik, 1989). We can regard the use of models as a means of coping with the complexity of the development process. A model reflects, through abstraction of detail, selected characteristics of the object, device or process in the real world that it stands for.