Search

CaseMaker is an interactive environment for intelligent case-authoring support in CREST, a case-based reasoner for estimation tasks, in which the selection of cases for addition to a case library is guided by empirical evaluation of the coverage contributions of candidate cases. I present a new version of the environment called CaseMaker-2 that is designed to support case authoring more effectively by eliminating a previous requirement for the evaluation of candidate cases to be repeated following the addition of a new case to the library. A key role in the approach is played by an algorithm for dynamic partitioning of the space of uncovered cases in such a way that the partition containing a given case represents the minimum additional coverage provided by its addition to the case library. [ABSTRACT FROM AUTHOR]

Preceptor is a shell system for building rule-based expert systems to support clinical decision-making. Its highly-structured knowledge base is designed to reflect the variety of decision-making activities in patient management. We describe Preceptor's knowledge base and dialogue structures, its inference strategies, and its comprehensive facilities for knowledge base creation and management. We report its application in a study of prognostic indices in stroke, and describe our approach to knowledge acquisition in this application. [ABSTRACT FROM AUTHOR]

α,β-unsaturated derivatives of 2-methoxy-2H-3,4-dihydropyran were synthesized. Stereochemical considerations of these heterocycles are briefly discussed. These compounds undergo a rapid first order reaction with water to produce glutaraldehyde, a potent microbial sterilant. Copolymers derived from these compounds and HEMA were found to provide excellent ex amples of clear, hard-sanitizing films; presumably by reaction with at mospheric moisture. Microbiological assays indicate that the release rates of the active agent from the polymers vary greatly depending on the hydrophilic ity of the comonomers. [ABSTRACT FROM PUBLISHER]

An algorithm for rule discovery in databases is described which is based on the reasoning strategies of human diagnosticians. It differs from other algorithms in its hypothesis-driven approach and primarily qualitative assessment of rule interest. Upper and lower bounds are established for the value of a quantitative measure used in the algorithm to rank rules of equal qualitative interest. An example based on consumer choices is used to illustrate the rule discovery process. [ABSTRACT FROM AUTHOR]