Reference: Musen, M. A. Generation of Visual Languages for Development of Knowledge-Based Systems. Plenum, New York, 1989.
Abstract: Computer programs that contain the knowledge of human experts and that offer advice on the basis of that knowledge (expert or knowledge-based systems) are assuming increasing importance in commercial and industrial settings. From programs that configure complex electronic instruments, to programs that supervise oil drilling, to programs that perform risk analysis for insurance underwriters, knowledge-based systems have been created for myriad application tasks. Such systems typically contain large amounts of application-specific exptertise encoded as a knowledge base. A generic, application-independent program ( an inference engine) uses the knowledge base to generate situation specific recommendations. Expert systems are large computer programs. Thus, software-engineering principles that apply to the construction of conventional computer programs also should apply to the development of expert-system knowledge bases. Yet the computer scientists who build exptert systems (knowledge engineers) view knowledge acquisition-the process of interviewing application experts and encoding their expertise in machine-understandable format-as a problem that is qualitatively different from that of standard programming. In this chapter, I explore some of the difficulties of knowledge acquisition, and demonstrate how the use of visual languages can ease the development of certain classes of expert sytems. I concentrate on PROTEGE, a tool that constructs visual languages that are tailored for specific knowledge acquisition tasks. First, however, it is necessary to describe the problem of knowledge acquisition in more detail.