Systematic introduction to expert systems. Knowledge representations and problem-solving methods

From MaRDI portal





The aim of this introduction is to give a systematic review of the state of the art with regard to the identification and realization of strong problem-solving methods and to indicate the prospects for their integration. It is based on experience gained by the author in consultations with users in numerous knowledge engineering courses and individual discussions, in which the main topic was usually the identification of problem-solving methods and knowledge representations in the different domains, on the results of our development and testing of tools for some of the problem-solving methods and for graphical knowledge acquisition techniques, and on an analysis of the relevant literature, which makes no claim to be exhaustive, however, on account of the great variety of more or less strong problem-solving methods and the rapid development of the field. Part I contains a general introduction and a characterization of expert systems. The main feature of expert systems is the clear separation which is made between problem-solving strategies and knowledge. It is this separation which enables the knowledge contents to be changed easily and the solutions to be explained by indicating the knowledge used for their derivation. Part II describes the basic techniques of knowledge representation commonly used in expert systems and the associated inference strategies: logic as the reference standard for the other techniques, rules with forward and backward chaining, frames with inheritance and attached procedures, constraints with local propagation, probabilistic reasoning with evidence values, non-monotonic reasoning with dependency networks, and temporal reasoning with temporal data bases. In Part III we discuss various approaches to the division of problems into classes, give a review of our own scheme and describe the uniform basic structure for the representation of the problem-solving methods. In Parts IV, V, and VI the main problem classes diagnostics (classification), construction and simulation are treated in detail. First their domains, problem types, problem features and problem-solving methods are analyzed. Then, for each of the three to six problem-solving methods in each main class, the basic idea, the object types and their attribute structures, the basic methods of knowledge manipulation, the graphical primitives for knowledge acquisition, the state of the art, and as far as possible an example of a corresponding tool from our research group are given. Finally, in Part VII, some aspects of the integration of problem-solving methods are discussed and illustrated with examples of their application.




Cited in
(31)


Describes a project that uses

Uses Software






This page was built for publication: Systematic introduction to expert systems. Knowledge representations and problem-solving methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4277012)