Automatic sketch map creation from labeled planar graph

Abstract

Maps constructed in Euclidean space are commonly used to visually present information about the real world. However, their creation is resource intensive, be it financial, technical, human, or time-consuming, which can limit their timeliness and detail. A simpler form of visualization of data about the real world is represented by sketch maps, which mainly capture the topology and mutual features’ spatial location. By default, they are drawn by hand. This presupposes that the creator has a good knowledge of the depicted territory, can create a cognitive map, and is skilled in transforming it into a graphical form. Sketch maps can be detailed and up-to-date if these prerequisites are met. Our question was whether it is possible to meet these assumptions in another way: acquire knowledge of the territory by processing narratives related to the area of interest, create a suitable computer representation for further processing, and automatically generate the resulting sketch map. This article presents the last step – creating a sketch map based on spatial data acquired from narratives. The results show that even without metric data, it is possible to automatically generate a sketch map visually close to the actual situation.