An expert system for journal selection

The National Library of Medicine indexes about 3000 of the world´s 20000 biomedical journals and maintains various indices to health-professional literature. The decision on which journals to index is based on a policy that is difficult to apply precisely. The current method for selecting journals is described and experiments in building an expert system to facilitate that task are discussed. It is explained how knowledge refinement strategies developed for a medical-diagnosis expert system were adapted to the present task, as well as how the massive database and knowledge base maintained by the library was tapped to support the journal selection decision-making process. The application of a learning program to the expert system to improve its accuracy is discussed.<>