Showing papers on "Domain knowledge published in 1980"

A domain-independent system that aids in constructing knowledge-based consultation programs

Abstract: : This thesis demonstrates an effective domain-independent system, called EMYCIN, for constructing one class of expert computer programs: rule-based consultants. Such a consultant uses knowledge specific to a problem domain to provide consultative advice to a client. Domain knowledge is represented in EMYCIN primarily as production rules, which are applied by a goal-directed backward-chaining control structure. Rules and consultation data may have associated measures of certainty, and incomplete data are allowed. The system includes an explanation facility that can display the line of reasoning followed by the consultation program, or answer questions from the client about the contents of its knowledge base. Other built-in human-engineering features allow the system architect to produce, with a minimum of effort, a consultation program that is pleasing in appearance to the client. To aid the system designer in producing a knowledge base for a domain quickly and accurately, EMYCIN provides a terse, stylized, but easily understood language for writing rules; performs extensive checks to catch common user errors, such as misspellings; and handles all necessary bookkeeping chores. To improve efficiency in a running consultation program, EMYCIN provides a rule compiler that transforms the system's production rules into a decision tree, eliminating the redundant computation inherent in a rule interpreter. It then compiles the resulting tree into machine code. The program can thereby use an efficient deductive mechanism for running the actual consultation, while the flexible rule format remains available for acquisition, explanation, and debugging.

Knowledge-based simulation

Abstract: Knowledge engineering has been successfully applied in many domains to create knowledge-based "expert" systems. We have applied this technology to the area of large-scale simulation and have implemented ROSS, a Rule-Oriented Simulation System, that simulates military air battles. Alternative decision-making behaviors have been extracted from experts and encoded as object-oriented rules. Browsing of the knowledge and explanation of events occur at various levels of abstraction.

Prototypes and production rules: a knowledge representation for computer consultations

Abstract: : This thesis presents a system called CENTAUR, which demonstrates the effectiveness of representing prototypical knowledge in a combination of frames and production rules for performing computer consultations. Key knowledge representation and control structure problems in production rule systems similar to MYCIN are identified, and a set of important characteristics of the structures used for representing problem-solving knowledge is given. CENTAUR's frames, or prototypes, complement the production rules to satisfy these characteristics and represent expected patterns of data that permit a more focused, hypothesis-directed approach to problem solving. Among the characteristics identified as desirable in the representation structures are the ability to explicitly represent (a) prototypical cases, (b) the context in which knowledge is applied, and (c) the strategies for applying that knowledge. CENTAUR's prototypes consist of patterns of knowledge in the domain which serve as broad contexts, guiding the more detailed processing of the production rules. Strategies for the consultation, or control knowledge, are represented in the prototypes separately from other kinds of domain knowledge. This allows the domain expert to specify control knowledge that is specific to each prototype. Examples are presented which demonstrate how this explicit representation facilitates explanations of the system's reasoning. Further, the organization of knowledge in CENTAUR provides a useful framework for acquiring new knowledge.

Advice-Taking and Knowledge Refinement: An Iterative View of Skill Acquisition

Abstract: : This paper discusses skill development as an iterative process that coverts advice into plans and, ultimately, converts these plans into behaviors An overall model is summarized While this framework treats learning as a largely domain-independent enterprise, it motivates two caveats First, we believe every skill is largely domain-dependent Whatever domain independence exists is attributable to the general skills that underlie initial skill acquisition and subsequent skill improvement Initial skill acquisition depends on the general and complex advice-taking skills of understanding and knowledge programming In this paper, we have developed many aspects of the advice-taking process The second phase of learning also employs numerous and relatively general skills In this phase, diagnostic and learning rules identify and rectify erroneous bits of knowledge The second caveat on domain-independence recognizes the important role that domain knowledge plays in diagnosis and refinement A learner's ability to apply diagnosis and learning rules will also depend on his or her familiarity with and expertise in the problem domain Although these heuristic and learning rules are domain-independent, to apply these rules a learner must be able to reason deductively about and with the entailments of his or her domain knowledge

Computer Science and the Representation of Knowledge

Abstract: This chapter contains sections titled: Parts and Wholes, Old and New Concepts of Knowledge and Intelligence, Research in Artificial Intelligence, Conceptual Representations, Self-Knowledge and Programming Knowledge, Editors' Postscript, References

Representation of control knowledge in expert systems

Abstract: This paper presents the results of research done on the representation of control knowledge in rule-based expert systems." It discusses the problems of representing control knowledge implicitly in object-level inference rules and presents specific examples from a MYCIN-like consultation system called PUFF. As an alternative, the explicit representation of control knowledge in slots of a frame-like data structure is demonstrated in the CENTAUR system. Explicit representation of control knowledge has significant advantages both for the acquisition and modification of domain knowledge and for explanations of how knowledge is used in the expert system.