The process of topographic map production at the national mapping agencies is undergoing steady change. The pressure to reduce production cost in particular hassled to the development and use of new approaches. Traditionally, separate datasets and models have been used to produce topographic maps of different scales and even the sources of data often differ. As a consequence, the derived datasets on different scales are usually inconsistent and thus cannot be updated in a common process. With the transition to GIS-based cartography, these disadvantages will be overcome. The research and development presented here have been carried out within an application-oriented research project of the University of Zurich in cooperation with the company Axes Systems. The focus has been on improving automated generalisation for topographic map production. A multiple representation database with extended linkage information was first implemented. These so-called vertical relations enable the storage of meta-information, which is needed to make the generalisations used during incremental updates. Second, a data model of horizontal relations was developed, to model contextual information explicitly. Examples of applications are the subdivision and partition of datasets, and the modelling of feature groups. Third, a service-oriented architecture was established, enabling a platform-independent development and the provision of generalisation functionality. In addition, an embedded workflow management system was introduced, consisting of a workflow editor and workflow engine, to support the interactive configuration of the generalisation process and the automated execution of generalisation services.