Database Reference
In-Depth Information
unlikely to be valid for the prison. It might be that we choose to model the knowledge
as a single Building, with a current and a former use, but what if the building has
been extended to make it into a hotel? Then, the spatial footprint will be different.
In one context, the building is still the same but now has a different use and a dif-
ferent spatial footprint, but in another context, it can be thought of as two separate
things. There is no easy answer to this knowledge modeling problem—we just point
it out here to warn the reader that spatial collocation is not sufficient to establish
equivalence, and while RDF offers greater opportunities for modeling the nuances
of our knowledge, it also can create deeper pitfalls when equivalence relations are
applied without due care and attention.
A GIS specialist is likely to be familiar with hierarchical relationships, equivalence
or mereological (“part of”) relationships; however, it is important when modeling
data using RDFS to consider using other relationships that may well express your
knowledge more accurately and not feel restricted to just hierarchical or mereological
ones. For example, a Post Office could be “within” a Shopping Mall if it is located
spatially inside the footprint of the Shopping Mall, rather than necessarily be consid-
ered “part of” it, and the two should certainly not be considered to be related in an
explicit hierarchy. Since hierarchical relationships result in the child inheriting all the
parent's properties, the “is a kind of” relationship of RDFS needs to be used with care.
Remember that
any
arbitrary relationship can be created in RDF; hierarchy is not the
only form of relationship!
RDF can offer some advantages to the geographer in modeling vagueness. For
example, if a boundary cannot be precisely measured, a semantic term like “nearby”
may be more appropriate for use as an RDF relationship than attempting to quantify
vague boundaries explicitly and numerically. Alternatively, a vocabulary like SKOS
could be used to indicate an unspecified similarity between two concepts.
5.9 SUMMARY
This chapter has explained how to model data in RDF using statements of triples
and covered the basics of the RDF language, looking at the main varieties you might
encounter: RDF/XML, Turtle, N Triples, and RDFa. We have also talked about how
to use RDFS to specify which classes and properties you intend to use to struc-
ture your data in RDF. We covered some basic RDF vocabularies like Dublin Core,
FOAF, and Geo and pointed out some potential traps to avoid as a geographer com-
ing new to RDF data modeling. In the next chapter, armed with this grounding of
RDF and basic knowledge representation in the triple structure, we explain how to
organize GI as Linked Data.
NOTES
1. RDFS is a simple ontology language.
3. Before you go looking for Mereashire in GeoNames, please note this is a fictional example.
4. For example,
http://rdfabout.com/demo/validator/
.
