Knowledge Acquisition for Generalization Rules

How to replace the human element of generalization with computer algorithms or rules in knowledge bases has been frequently discussed, but we have not succeeded in formulating these rules very well. The purpose of this article is to illustrate the problem of knowledge acquisition for generalization of topographic maps. Three studies to derive rule-based knowledge for automatic map generalization are presented and analyzed. In the tests, cartographers were asked to interpret map objects to be generalized and to describe the basis of their decisions. The studies showed that by interviewing cartographers, much of the domain knowledge can be gathered, but the most time-consuming part of documenting this knowledge is to analyze the data and formalize the results. It was also found that there is important domain knowledge on generalization that has not previously been documented. The tests performed resulted in the discovery of four categories of declarative rules: geometric, topological, context-related, and culture-related rules.