Cartograms of self-organizing maps to explore user-generated content
André Bruggmann, Marco M Salvini, Sara Irina Fabrikant · Zurich Open Repository and Archive (University of Zurich) · 2013
As the amount of user-generated content dramatically rises, the need to structure these data, to extract relevant semantic relationships buried in the data, and to visualize found relationships appropriately has significantly risen as well.We suggest an innovative method to structure, visualize, and visually explore user-generated data using a cartogram of a self-organizing map.This distorted self-organizing map overcomes the cognitive limitations of the traditional self-organizing map by combining this neural network mapping approach with cartographic methods used to generate cartograms.First, our novel mapping approach is put to a rigorous test in a case study aimed to uncover the latent semantic structure from text documents in the Wikipedia Encyclopedia.Second, the latent structure uncovered with the self-organizing map cartogram is systematically evaluated by comparing it to an established network visualization method and output.The resulting self-organizing map cartogram reveals relevant structures in the considered Wikipedia data.The comparative evaluation confirms the validity and the stability of the found patterns, and therefore of our novel visualization solution.This paper further contributes to the spatialization research line, by expanding the use of well-established and empirically evaluated cartographic depiction methods to the visualization of non-geographic data, such as, for example, user-generated data increasingly available in today's networked information society.