Knowledge Acquisition最新文献

Architectures and methodologies provide knowledge acquisition researchers and practitioners broad working frameworks within which to explore a variety of questions. However, in order to pursue the development of robust reasoning systems rigorously with reasonable expectations of the development process, empirical assessments of the knowledge acquisition process and of heuristic inputs to the process are essential. In this paper we discuss an experiment whose results support the use of a group knowledge acquisition technique in combination with other knowledge acquisition techniques.

Metaphor is frequently used to describe the abstract aspects of science in scientific fields. Authors' choice of mataphor reflects and affects their thoughts, and through these, their practices in the field. For this reason, we propose that metaphors are appropriate to understanding approaches to knowledge acquisition. Salmond, an anthropologist, provides a methodological model of how metaphor in text can be manually analysed and used to classify the text according to its perspective on the phenomena of study. She picks anthropological references for her application of the method. We expand upon Salmond's method of metaphorical analysis and apply it to key concepts, such as the nature of knowledge, in: (1) text on knowledge acquisition; and (2) text on ethnomethodology, a sociological approach which has recently been applied to knowledge acquisition. For instance, the knowledge acquistion text was found to claim that knowledge is a human construction/tool, while the enthnomethodology text communicated that knowledge is rooted in social activity. We discuss and compare the results of our metaphoric analyses of the knowledge acquistion and ethnomethodology text, as well as those from Salmond's analyses of the anthropological text. We develop Salmond's ideas for classifying approaches and propose a three-part classification scheme: task-, concept- and domain-based. The perspective and application of each classification type are described. We illustrate how to use this classification scheme by classifying the enthnomethodology approach.

There is much current interest in automation of manual elicitation techniques, but little is known about whether automated versions of a technique produce similar results to the manual versions. This paper describes a formal comparison between an item sort, a card sort and a computerized label sort in the same domain. No significant differences were found between the types of knowledge elicited by different types of sort. These findings suggest that computerized implementations of sorting procedures will elicit the same knowledge as manual sorts.

This result also emphasizes the need for advice about knowledge elicitation to be based on formal experimental results rather than on assumptions, a-priori reasoning or case studies.

The Knowledge Acquisition Support Environment (KASE) and some case studies on expert system development are described. DiPROS is a diagnostic expert system building tool, which has a knowledge base editor and an inference engine specialized for diagnostic tasks. DiKAST is an interactive knowledge-acquisition (KA) tool, which guides repairs on defects in the knowledge base. In order to build this KA facility, many kinds of KA interview strategies were formalized. We applied KASE tools in building a defect diagnosis system for a color picture tube manufacturing plant. The prototype system was able to be developed in a short time by the domain experts alone, who had no previous experience in computer programming.

When acquiring knowledge from multiple experts, one has to be aware of the fact that differences of opinion among them may exist. A procedure has been devised that allows for a structured comparison of ideas of multiple experts. Perceived similarities and dissimilarities between statistical analysis methods have been elicited using a card-sorting technique based on the Repertory Grid. Both the elements to be judged and the constructs generated could be chosen freely by each individual expert. Grids were compared by transforming them through principal components analyses into configurations. These configurations have been matched using Procrustes analysis and dimension weighting. This analysis procedure uncovers systematic differences in perceptions, that would otherwise have gone unnoticed.

For the purpose of modelling commonsense reasoning, we investigate connectionist models of rule-based reasoning, and show that while such models can usually carry out reasoning in exactly the same way as symbolic systems, they have more to offer in terms of commonsense reasoning. A connectionist architecture, CONSYDERR, is proposed for capturing certain commonsense reasoning competence, which partially remedies the brittleness problem in traditional rule-based systems. The architecture employs a two-level, dual representational scheme, which utilizes both localist and distributed representations and explores the synergy resulting from the interaction between the two. CONSYDERR is therefore capable of accounting for many difficult patterns in commonsense reasoning with this simple combination of the two levels. This work shows that connectionist models of reasoning are not just “implementations” of their symbolic counterparts, but better computational models of commonsense reasoning.