Abstract: | Successful implementation of an intelligent system for automated map generalization requires formalization of cartographic principles that are, in many cases, only intuitively understood. Formalizing these principles requires acquisition and re-expression in the form of semantic nets, frames, production rules, or similar formalization methods. The various techniques for cartographic knowledge acquisition have been discussed on a theoretical basis; however, little empirical research has been conducted. This paper reports on empirical acquisition of cartographic knowledge by reverse engineering; that is, on trying to recapitulate decisions made on published documents or maps. The work is based on a computer-assisted multi-scale inventory of the Austrian National Topographic Map Series. Queries of the relational database, within which inventory data are stored, lead to the formulation of prototype production rules for modifying map symbols during automatic scale changes. Components of map generalization expressed in such rules include the selection behavior of settlement, transportation, and hydrographic objects, and the degree of simplification of settlement domains and building clusters. The acquired cartographic knowledge reveals quantitative relations between map elements and the changes in these relations that occur with scale transition. These insights can guide subsequent knowledge refinement using other acquisition methods. This paper provides, in addition, a conceptual framework by which other topographic map series may be compared at multiple scales. |