Abstract

Knowledge engineering promises to extend the range of tasks that can be automated and to make information systems more accessible to end-users such as policy makers. It can make knowledge more widely available and improve its consistency. Neural networks could be used for pattern recognition and data analysis tasks, machine learning techniques can alleviate the acquisition and maintenance of knowledge bases. This spring the Informatics Service Center has started a project to explore these techniques and promote their use in promising areas. At present the use of these techniques at the RIVM is still limited. This report describes projects in which use of these techniques was attempted or contemplated. Apart from a number of exploratory studies, knowledge engineering is used in two systems, a toxicological information system for physicians and a support system for the selection of measures to obtain reduced emission levels. Neural networks are used for the interpretation of infrared spectra of ground samples. Given the mission of the RIVM as a knowledge broker and given the growing integration of knowledge engineering and main-stream information technology and the promise of new technologies such as automated induction and neural networks, there is every reason actively to pursue the use of these techniques at the RIVM. Potential application areas will be further elaborated in future reports.