Abstract
LINKED DATArepresent a huge source of information and knowledge both inside and outside the Semantic Web research field. However, a broad and extensive adoption of this technology is still prevented by the difficulties that users find when approaching these large and complexly structured sources of data. Typically, even the experts of this field are disoriented in understanding both the main structure and the details characterizing how these data have been modelled. Filling this initial gap means performing very meticulous, costly and time-consuming analysis by using a specific query language.
A possible way for solving the problem is adopting other approaches that are more centered on the use of visual representations, leveraging the human visual perceptual channel. Differently from the ordinary methods widely investigated in the literature, we propose an approach based on Information Visualization techniques and cartographic principles. Large and complex data are exactly the kind of information Cartography has been dealing with for centuries. This capability in representing data can be fur- ther augmented by the interactive mechanisms that can be implemented using modern computer-based solutions.
Navigating geographic spaces and effectively observing their information is a task humans are quite used to. The approach presented in this thesis produces abstract maps resembling the traditional geographic ones and thus allows users to reuse their cognitive skills and prior knowledge in reading maps for the navigation of abstract map-like visualizations of Linked Data sets. In order to produce these particular visualizations, a specific process called spatialization has been employed to assign data a range of spatial attributes, such as size, shape and position on a two-dimensional surface. The specific spatialization we used is based on space-filling curves, fractal curves having the feature of entirely filling the space. In particular, by exploiting the fractal nature of these curves, a novel technique for properly expressing data and efficiently spatialize them in a scalable way has been devised.