Publications
Use Of An E-R Diagram In The Design Of A Feature Dictionary
For A Multi-Domain Medical Knowledge Base
11/01/1987
F. Naeymi-Rad1, T. Koschmann2, R Rosenthal1, D. Trace1, S. Naeymi-Rad1,
J. Swanson1, C Lee1, R Carlson3, and M.H. Weil1, M. Evens3
1University of Health Sciences/The Chicago Medical School, North Chicago, IL 60004
2Xerox Al Systems, Des Plaines, IL 60013
3Illinois Institute of Technology, Department of Computer Science, Chicago, IL 60616
Abstract
Past research has shown that the performance of a Bayesian system is crucially dependent on a consistent set of features. Because different terminology is often used by physicians of difficult specialties or locations to refer to the same feature, it is necessary that the knowledge base provide a means to compensate for discrepancies in terminology or definition. For this purpose, a new dale base called the Feature Dictionary has been designed with an E-R model. This dictionary will provide a means for transcending the medical "language barrier" by recognizing the entry of a feature into the knowledge base no matter which of the common terms the particular expert has used. For instances in which the expert comes across an unfamiliar term in the system, he or she will be able to call up a definition of up to four lines in length which will describe exactly what the feature is to which the term refers. The feature dictionary would enable us to merge existing MEDAS databases, evaluate and compare knowledge bases, or translate between MEDAS knowledge bases and medical knowledge bases of different types. II could be used in educating health care professionals and in collecting and analyzing dale for epidemiological research independent of the MEDAS system. The Feature Dictionary will be used in conjunction with the MEDAS (Medical Emergency Decision Assistance System) knowledge engineering tool TOOLBOX. The MEDAS TOOLBOX has been used in the design of diagnostic consultants for a variety of medical domains.
Download this publication as a PDF file