Display of Large Amounts of Geometries

Some spatial RDF datasets, such as DBpedia or Freebase, contain significantly more spatial information than can be reasonably retrieved and displayed on a map in a web application considering bandwidth and performance restrictions. Facete handles such cases using a quad tree data structure:

Based on the users constraints on the facets and the geo-link, a corresponding SPARQL concept, named geo-concept, is created. The geo-concept specifies the set of resources to be shown on the map.

A count of the number of instances matching the geo-concept is requested. If the count is below a configured threshold, all instances are retrieved at once and placed into the root node of the quad tree.

If this count exceeds the threshold, the extent of the whole map is split recursively into four tiles of equal size. The recursion stops if either a maximum depth is reached, or if the tiles have reached a certain relative size when compared to the map viewport (e.g. when about 4 × 4 tiles are visible). For each tile, the geo-concept is then modified to only refer to resources within that tiles' bounding box. A tile's resources are only retrieved, if the new count is again below a configured threshold.

Tiles that still contain too many geometries are rendered as boxes on the map.

An example of such display is shown in Fig. 7, which shows a subset of the approx. 20.000 resources with geo-positions in Germany. For each set of constraints, Facete creates a new quad tree that acts as a cache for the user's current configuration.

Fig. 7. Display of Freebase instances in Germany.

Related Work

The RelFinder system [6] is capable of finding property paths connecting a pair of resources, whereas Facete searches for paths between SPARQL concepts. Over the past decade, faceted search has become omnipresent, such as in web shop and content management systems. Apache Solr [1] is a popular system that offers extensive faceted search features, however, it does not offer native SPARQL support and thus requires pre-processing of RDF data. Rhizomer [1] and the Virtuoso faceted browser [2] support changing the focus from one set of resources to a related one (known as pivoting). However, with these systems, the user actually navigates between related list views of resources, whereas in Facete the user pins facets as new columns to the table view.

  • [1] lucene.apache.org/solr/
  • [2] virtuoso.openlinksw.com/dataspace/dav/wiki/Main/ VirtuosoFacetsWebService
< Prev   CONTENTS   Next >