This application is connected to a newly financed eSSENCE project with the same name at Lund University.
Traditional automated map labeling methods are based on quantification of cartographic rules and optimisation techniques. However, these methods are not good enough for high-quality maps which implies that a substantial part of the map labelling is still made interactively. In this project we use a data-driven approach to automated map labeling. By a cooperation with the Swedish company T-Kartor we have access to around 50,000 unique map samples where the final placement of the labels has been performed interactively by a cartographer. These map examples are utilized both for improving the current map labeling algorithms and for selecting candidate solutions. The latter step is performed by training an AI-network.