Reference: Washington, R.; Boureau, L.; & Hayes-Roth, B. Using Knowledge for Real-Time Input Data Management. KSL, March, 1990.
Abstract: A complex, real-time domain produces more data than an AI reasoning system can process. Thus some sort of preprocessing and filtering of the data must occur. To produce appropriate data for the reasoning system, the preprocessor's selection criteria should accurately reflect the reasoning systems knowledge as it applies to data selection. In addition, the preprocessor should give the reasoning system information about the data that will help the system effectively reason with the data. We have developed a method of knowledge-based data management, perceptual preprocessing, that effectively preprocesses the data for an AI system in a real-time domain. Furthermore, we have implemented this in the context of a real-time knowledge- based system for monitoring a surgical intensive-care unit.