DBpedia Map - First-level classes of the ontology (with links)

A first step towards the creation of a map of DBpedia. This experiment shows the first-level classes of the DBpedia ontology and the links between them. You can zoom in, hover and click on links and nodes to obtain more detailed information.

A node’s area grows linearly with the amount of entities of the relative class, while link thickness is representative of the amount of triples that have entities of the corresponding classes as subject and object (the encoding is nonlinear to make the diagram readable - so the thickness is exaggerated). Directionality is conveyed with animation (as in this example).

Because hierarchical links are not depicted, this is an implicit visualization (Schulz et al. 2010) of the ontology tree. Only the first level is shown for the moment - deeper nodes will be drawn inside their parents (circular treemap). At the same time, it is an explicit visualization of the graph defined by the triples. We decided to trade some compactness (treemaps’ best feature) with plenty of room to show the important information carried by links. Moreover, because links are bundled and attracted to the center, the hierarchy is more explicitly shown.

Many issues are still open:

• Zooming into a node should reveal its inner structure.
• The relative scale of nodes should not be too large, to avoid to give the feeling of navigating a huge, mostly empty space.
• The map can either have a coherent node scale in the whole space, or have vast empty spaces. This is due to the fact that the sum of the areas of children nodes has to be much lower than the area of the parent, because we saved a lot of space for depicting links.
• Nodes should be color-coded as in previous experiments.
• We have to think what to do exactly with the “other” class (a dummy class we used to merge many first-level classes together)
• Ingoing and outgoing links should form two parallel “roads”, rather than overlap. This is a nontrivial problem, because offsetting a bezier curve is not easy.