Geo Werkgroep PLDN


Wouter Beek (w.g.j.beek@vu.nl, wouter@triply.cc)

Properties of Cadastral Big Data

  • Over 5,000,000,000 graph edges
  • Over 500,000,000 geo shapes
  • High precision
  • Performance

Challenging: Geo + Linked Data

GeoSPARQL

Components:

  • Core
  • Topology Vocabulary (Relation Family)
  • Geometry (Serialization)
  • Geometry Topology (Relation Family, Serialization)
  • RDFS Entailment
  • Query Rewrite

Relation Families:

Simple Features, Egenhofer, RCC8

PrecisionMeaning
+/-east/west or north/south
1.0e2must be longitude
1.0e11111 kilometer, continent
1.0e0111 kilometer, country
1.0e-111 kilometer, city
1.0e-21.1 kilometer, village
1.0e-3111 meter, large field
1.0e-411 meter, parcel of land
1.0e-51.1 meter, tree
1.0e-611 centimeter, engineering
1.0e-711 milimeter
1.0e-81.1 milimeter, tectonic plate motion
1.0e-9110 micron, microscopy

Implementations

Implementations

  1. Not that many implementations to choose from.
  2. FOSS implementations do not perform.
  3. Challenging to deliver Open endpoints.

SPARQL Benchmarks

Synthetic:

  • SP2Bench
  • WatDiv
  • BSBM (Berlin SPARQL Benchmark)
  • LUBM (Lehigh University Benchmark)

Real-life:

  • DBPSB (DBpedia SPARQL Benchmark)
  • FEASIBLE

GeoSPARQL Benchmarks

  • Bellini2015: Benchmarking RDF Stores for Smart City Services
  • Geographica (2013, small data)
  • Volume: number & detail
  • Variety: large & small
  • Veracity: differences in precision

Open Data

Use (Geo)SPARQL endpoints i.o. copying data locally.