Database Reference
In-Depth Information
County be classed as a Physical Region or a Political Geographic Object? Is a flood an
endurant or a perdurant? It depends on your point of view. These quandaries become
even more apparent when confronted with terms like “Non-Agentive Social Object”
or “Abstract.” The domain expert can spend more time puzzling over where to place
a domain concept in the upper ontology hierarchy than actually building his or her
own ontology. Furthermore, upper ontologies are inherently based on the principle
of centralization: If all our domain ontologies are based on the same upper ontology,
it will be easy to integrate them. However, as we have mentioned, the Web is inher-
ently decentralized, and there is more than one upper ontology to choose from, so in
this topic we would not particularly recommend that the GI domain expert should
use an upper ontology as a base when building an ontology. Upper ontologies are
useful to study as they present good knowledge modeling discipline when consider-
ing the more difficult to model concepts like matter (“Sand” or “Rock”) or the dif-
ference between Constitution and Identity (e.g., a stream is a body of water, but even
if it dries up during summer it is still a stream).
In the same vein as upper ontologies, in the early days of the Semantic Web several
large-scale domain ontologies went into development, for example, the very success-
ful GALEN ontology for the biomedical domain (Rector, Rogers, and Pole, 1996).
These ontologies tended to consist of hundreds of thousands of concepts, usually in a
scientific field, and were authored by committee. In the GI domain, ontological devel-
opment was influenced by philosophy and a desire to develop a “complete ontology
of the geospatial world [that] would need to comprehend not only the common-sense
world of primary theory but also the field-based ontologies that are used to model
runoff and erosion” (Smith and Mark, 2003). Authoring ontologies is discussed fur-
ther in Chapter 10, but for now, we can just note that, almost ten years later, this holy
grail of a single complete ontology of the geospatial world does not exist, and there is
a persuasive argument to say that one monolithic ontology is not necessary or obtain-
able. Instead, smaller, more agile, rapidly developed ontologies describing specific
domains within geography for specific purposes (e.g., the Ordnance Survey Spatial
Relations Ontology
6
) have been more successfully deployed.
It was perhaps the concentration of the Semantic Web research community on
developing large ontologies and optimizing ever-more-complex reasoners that led
some to rethink the direction in which the Semantic Web was heading. The Linked
Data movement started with a return to the drawing board to concentrate on expos-
ing data to the Web that was hidden in proprietary databases, structured in myriad
ways. To do this, they recommended structuring data in a standard format: RDF,
which could also be used to specify links into and out of each dataset.
As described in Heath and Bizer's topic on the subject (Heath and Bizer, 2011),
“Linked Data provides a publishing paradigm in which not only documents, but
also data, can be a first class citizen of the Web, thereby enabling the extension
of the Web with a global data space based on open standards—the Web of Data.”
This means, therefore, that a GI specialist wishing to share data on the Web could
publish it as RDF and include links to other datasets to allow the data to be discov-
erable by Web crawlers. Since geography is so often used as a backdrop to other
information sources, it is frequently used in the Web of Data to link together other
RDF datasets that have a geographic element. This is apparent in the picture of the
