Knowledge Acquisition最新文献

Rapid prototyping and tool reusability have pushed knowledge acquisition research to investigate method-specific knowledge acquisition tools appropriate for predetermined problem-solving methods. We believe that method-dependent knowledge acquisition is not the only approach. The aim of our research is to develop powerful yet versatile machine learning mechanisms that can be incorporated into general-purpose but practical knowledge acquisition tools. This paper shows through examples the practical advantages of this approach. In particular, we illustrate how existing knowledge can be used to facilitate knowledge acquisition through analogy mechanisms within a domain and across domains. Our sample knowledge acquisition dialogues with a domain expert illustrate which parts of the process are addressed by the human and which parts are automated by the tool, in a synergistic cooperation for knowledge-base extension and refinement. The paper also describes briefly the EXPECT problem-solving architecture that facilitates this approach to knowledge acquisition.

An Active Expert methodology towards knowledge acquisition is proposed. Briefly this methodology implies that the expert should take as active a part as possible in the creation of the knowledge base. The knowledge engineer should act more like a teacher of knowledge structuring, as a tool designer and as a catalyst in the dialogue between the expert and the end-users. By doing so, many of the well-known problems with inter-human conflicts, knowledge engineer filtering, expert and end-user acceptance and maintenance could be reduced. The methodology has been developed during a 10-year period with three practical projects and a close cooperation with research in tool-based knowledge acquisition as the main empirical material. A major part of the paper is devoted to a description of the Active Expert methodology divided into 10 phases. Each phase is exemplified with material from practical projects.

A general approach to knowledge elicitation in interactive learning systems is presented which both improves a knowledge base by removing inconsistencies and extends the representation space for learning. This approach addresses the problem of learning "new terms" with interactive learning systems. Two methods that illustrate this approach are implemented in the learning apprentice system NeoDISCIPLE, using a concept-based representation that is very appropriate for learning. At the same time, the representation facilitates knowledge elicitation associated with human-oriented representations like, for instance, repertory grids. Both methods are consistency-driven in that they elicit knowledge from a human expert in order to remove inconsistencies in the knowledge pieces learned by NeoDISCIPLE. The input to these methods is an inconsistent rule learned by NeoDISCIPLE, together with the examples from which the rule has been learned. The elicitation process is characterized by a guided interaction with the human expert, who is asked to make relevant distinctions pertaining to concepts appearing in the positive and negative examples of the rule. The first method elicits concept properties through a goal-driven property transfer from one concept to another, and the second one elicits concepts using a goal-driven conceptual clustering. In both cases the elicited knowledge is used to improve the inconsistent rule while simultaneously extending the representation space for learning.

Knowledge acquisition is a modeling activity for the development of knowledge-based systems. Developers can take advantage of domain-specific knowledge-acquisition tools for the construction of knowledge bases. Modeling for the development of such tools is an activity that is analogous to the modeling for knowledge-based systems. However, there are important differences between the appropriate models for these two goals. The models for the development of domain-specific knowledge-acquisition tools involved can be divided into three major classes: knowledge-structure models, knowledge-acquisition models and design models. These models can help developers to create domain-specific tools for new application domains.

Many concept formation systems construct disjoint-concept trees. However, a priori imposed tree structures may restrict the application of these systems in some domains. A joint concept formation scheme is thus proposed, which learns from observation, and constructs acyclic directed concept graphs (trees are a special case). We show that the joint concept formation system can avoid or alleviate some problems the disjoint concept formation system would face, such as the unique winner and oscillation problems. We also demonstrate that a joint concept formation system is able to generate a concept tree if such a regularity is found among the data. The experimental results are consistent with the expectations that the joint system is a generalized version of the disjoint system and improves the learning performance. Joint concept formation extends the classic works, such as COBWEB and ARACHNE.