According to the authors, “this paper describes a knowledge-acquisition tool that builds expert systems for evaluating designs of electro-mechanical systems.” Although some tools for building and maintaining large knowledge bases have been developed in the last decade, building knowledge bases is still an art and often starts from scratch. The authors of this paper believe that a good understanding of the problem-solving method used is crucial. Their system is a demonstration of this approach.

The expert knowledge collected by KNACK is used for the evaluation of electromechanical systems for nuclear hardening. Two phases are defined to accomplish this task. Information gathering, for building a design description, is followed by evaluating the design and suggesting improvements.

The first expert system generated through KNACK, called a WRINGER, was also built in two consecutive steps. First, information was gathered on an electromechanical system in order to produce a design description in the form of a report, and, second, the design was evaluated and suggestions were made on how to improve it with respect to nuclear hardening.

Two problem-solving methods are used by the WRINGERs to determine the sequence of the subtasks to be solved as well as the organization of the knowledge base. The information gathering method is data-driven, and is based on the preliminary report structure given by the expert. Information gathering is done in different ways, such as asking direct questions, filling slots, and choosing from a menu. The WRINGER can also access a database and will try to deduce or compute what it needs before asking the user.

The method used for evaluation is, first, to apply consistency and completeness checks to the gathered information and then to evaluate the design for possible flaws. When flaws are found, the information gathering method is called again to get the missing information. The results of the evaluation are also given in report form.

KNACK, the knowledge acquisition tool that builds WRINGERs, which are written in OPS5, aids the expert in two ways: “(1) It elicits knowledge from the expert about a skeletal report, report fragments, and how to customize the report fragments; (2) it uses heuristics to infer additional knowledge.” To guide the user further, KNACK organizes knowledge into knowledge roles, which are characterized by templates.

This paper reports research that has progressed further. In a more recent paper, the authors have explained in a more complete way what report-driven means [1]. The area of investigation is considered important in the expert systems community and is now quite active. The references in the paper should be read first; however, more research is now under way that has not yet been published. The system described is interesting and seems quite powerful for freeing the expert from unnecessary work. The weakest point of the paper is that the examples are too short to make the knowledge-acquisition process clear. Reading the other paper may alleviate this problem.