Practical techniques for organizing and measuring knowledge.

Abstract

This research is concerned with the problem of making knowledge acquisition and representation practical for a wide variety of people. The key question investigated is the following: What features are needed in what this research defines as a knowledge management system, so that people who are not computer specialists can use it for tasks that involve manipulating complex ideas? In this research the needs of such users were evaluated, and several prototype systems were created, culminating in the creation of a system called CODE4. A key conclusion is as follows: Users need a tool that involves the synthesis of several techniques for organizing the knowledge. The research has led to ideas for two types of features: abstract knowledge representation features and user interface features. Some of the proposed abstract knowledge representation features that are novel or improved include: (1) a uniform treatment of units of knowledge which are called concepts; (2) the separation of the names of concepts from their unique identities so that concepts can have multiple names and a given name can be used for several concepts; (3) the classification of conceptual relations using property and statement hierarchies, and (4) facilities for smoothly integrating informal knowledge with knowledge that is more precisely represented. The above ideas are implemented in CODE4-KR, the abstract knowledge representation used in the CODE4 knowledge management system. Some of the user interface features that are found in CODE4, and are proposed for general use, are: (1) unlimited chained browsers; (2) knowledge maps, representing patterns of knowledge to be explored or processed; (3) fully interchangeable mediating representations operating on knowledge maps; (4) a uniform set of knowledge editing operations in the mediating representations, and (5) masks for highlighting and controlling the visibility of concepts. CODE4 has had a significant amount of serious use by users in such diverse fields as terminology, organizational modelling and software engineering. This has served as the basis for validating the above ideas about knowledge organization. To this end, several techniques have been used including: (1) a detailed questionnaire about CODE4 administered to many users, and (2) a static analysis of the knowledge bases using several newly-developed metrics. The metrics include seven largely independent measures of knowledge base complexity, as well as compound measures that allow the user to ascertain the completeness, well-formedness and information-richness of a knowledge base.