Database Reference
In-Depth Information
and having an explicit stated scope will help you avoid trying to describe the whole
world. The third initializing step in schema creation is to select what are known as
competency questions. These are in effect your test set: the questions that the RDFS
ontology and accompanying data must be able to answer correctly. Further in the
process, these will be translated into the RDF query language SPARQL, and the
Linked Data set queried, to make sure it fulfills your requirements.
If Merea Maps wants to create some Linked Data about its administrative regions,
it could state the purpose of its Linked Data to be: “To describe the administrative
geography of Merea and the topological relationships between them.” The scope
would be: “All levels of administrative area that occur in Merea, their sizes, point
locations, the topological relationships between areas of the same type. Authorities
that administer the regions are not included, nor are the spatial footprints of the
regions,” and they could have a set of competency questions, which would include,
among others:
Name all the Counties in Merea.
Find all the Cities in Merea.
Which Parishes are in Medina County?
What is the latitude/longitude of Medina City?
Find the Counties that border Medina County.
Find the Counties that border Lower Merea County.
While some ontologists advocate merely using competency questions as an indica-
tion of the kind of question the RDFS ontology should answer (Noy and McGuinness,
2001), we would instead advise that there should be as many competency questions
as necessary to cover all the queries you expect users of your data will want to ask.
This follows the principles of test-driven development, and by the time you have a
satisfactory list of competency questions, you will be well on your way to listing the
classes and property names need for your RDFS ontology, which comes in stage 2 of
the ontology design. The purpose, scope, and competency questions can be included
in an rdfs:comment in the Linked Data to document the RDFS ontology.
As we have just mentioned, the second part of deciding what your data is about
is to choose the class and property names. To do this, consider first the domain of
interest. What are the most important concepts in the domain, and how are they
linked together? For Merea Maps, the concepts will be things like Country, Region,
District, County, City, and Parish. These are not hierarchically related (e.g., a City is
not a kind of County), but they are related topologically, so a City will be in a County.
We need to think a bit more about the topological spatial relations now, as “in” is
quite a vague term. Merea Maps decides to use four topological relations:
Spatially Contains: The interior of one region completely contains the inte-
rior of the other region. Their boundaries may or may not intersect.
Spatially Contained In: The inverse of Spatially Contains. The interior of
one region is completely contained in the interior of the other region. Their
boundaries may or may not intersect.
Search WWH ::




Custom Search