Data-driven Expansion of Dense Regions - A Cartographic Approach in Literary Geography

Cartographic visualisation of the literary space is facing a major challenge resulting from different levels of detail within which the textual descriptions of settings are made by authors. Those range between very detailed descriptions within parts of a city to spatially spread events within a country, or long journeys across continents. Depending on the fictional texts, fictional action often concentrates on a few main places, resulting in high information density - in the form of various individual settings. As well, they are also embedded in a larger environment. When interactively analysing or choosing a section to print a literary map with individual spatial elements, the user has to choose a map scale, which will result in the output of either a detailed representation of a main place or the geographical overview of the fictional space having a small level of detail. However, it would be a great advantage to receive as much information as possible from one single map view. In order to achieve this, this paper presents a method that allows increasing the representable amount of information, depending on its density. Our method makes use of the diffusion algorithm that is used to create value-by-area cartograms. This algorithm is applied to an auxiliary density grid derived from the distribution of individual settings of a literary map that are to be displayed. The resulting distortion is subsequently transferred to the data of the literary map. Unlike the usual use of area cartograms, we do not aim to represent a statistical value; instead information density is used to provide space for the depiction of the information itself. Various parameters such as grid resolution, scale factor and a smoothing filter visually influence the final map distortion. On the basis of several map examples taken from the 'Literary Atlas of Europe' project database, distortion results generated from the application of different parameters are visually examined. Using this process, we improved the proposed approach and generalised an appropriate initial configuration for the automated generation of distorted maps through the usage of its information density.