Publications
Knowledge Generation via a Simple Grammar Supporting an Intelligent User Interface
11/06/1988
Lowell A. Carmony1, Frank Naeymi-Rad2, Robert Rosenthal2, Shon Naeymi-Rad2, David A. Trace2,
Eric Rackow2, Max Harry Weil2, and Martha Evens3
1Lake Forest College, Computer Science Department, Lake Forest, IL 60045
2UHS/The Chicago Medical School, North Chicago, IL 60064
3Illinois Institute of Technology, Chicago, IL 60616
Abstract
A dictionary of standard medical terms (called the Feature Dictionary), a grammar to control the format of features, and a standard portable file in which to archive patient data will permit the automatic comparison and evaluation of competing knowledge bases for MEDAS (Medical Emergency Decision Assistance System), as well as provide the user with an intelligent interface for the entry of patient data. The feature dictionary consists of simple binary features such as "Abdominal Pain", continuous-valued features such as "White Blood Count = 14,000", and derived or computed features such as "Blood Pressure = Systolic - Diastolic", but the medical expert describes knowledge to the system in terms of compound features such as "Sex = Female & Age > 2 & Hematocrit 37 to 42". The new system contains a grammar for compound and complex features and a run time parser to translate these features into reverse Polish notation. The parse trees are used to generate rules to support an intelligent user interface. Thus, the user need only set the binary features and enter the values for the continuous features and then the system at run time automatically sets the derived features as well as the range and compound features that are needed for MEDAS' Bayesian multi-membership Inference.
Download this publication as a PDF file