Canadian Forest Service Publications
Formation and use of intermediate inferences in advisory systems: a herbicide example. 1992. Thomson, A.J.; Williamson, D.R. AI Applications 6(4): 29-37.
Issued by: Pacific Forestry Centre
Catalog ID: 3286
CFS Availability: Order paper copy (free)
Focus on the knowledge used in product evaluation guides not only the structure of an expert system's knowledge base but also the inferencing procedure required to move from the original knowledge base to a final conclusion. Search procedures based on intermediate inferences, rather than on the original facts, are simpler and faster, and are more easily pruned, than are procedures based on the original knowledge base. Use of Prolog list structures to facilitate this process is demonstrated in an expert system for advising on selection of a forest herbicide. The structures in the original knowledge base used to represent basic herbicide properties were relatively large and complex. In the intermediate inferences, on the other hand, the structures were simplified and lists were often empty, and they represented a case-specific perspective on the "general knowledge" facts of the original knowledge base.