Database Reference
In-Depth Information
10.4.4.6 Breaking the Pattern Rules
Before concluding this discussion on patterns, we sometimes come across other
things where a pattern applies but we do not think they belong to the class that
defines the pattern. Merea Maps encounters this when describing Landform classes.
It starts to describe an Orchard:
Every Orchard is a kind of
Enclosed Land.
9
Every Orchard has part Trees.
Every Orchard is intended for
Agricultural Production.
Class: Orchard
SubClassOf: EnclosedLand, hasPart
some Tree, isIntendedFor some
AgriculturalProduction
and realizes that this looks very like the pattern created for Place. So, is an Orchard
really a Place and not a Landform, is it a Place
and
a Landform, or does Merea Maps
still consider it to be only a Landform? Once again, there is no absolute answer,
merely different solutions that may apply in different contexts. Strictly, Merea Maps
could argue that as an Orchard is somewhere where something was intended to hap-
pen—in this case growing fruit for agricultural purposes—then an Orchard must
be a Place. However, Merea Maps is also aware that the majority of its users would
naturally classify an Orchard as a Landform, not a Place; it also does not like the idea
of making it a subclass of both Place and Landform and so settles on describing it
as it has. The reader may disagree with this decision and choose a different solution.
This just highlights that all ontologies are imperfect compromises, and that there
are often situations for which there are no best solutions. It also demonstrates that
top-level classes such as Place and Landform can be really quite weak, with different
people having different ideas over membership.
10.4.5 p
roperties
We have so far largely discussed the construction of the ontology from a concept-
or class-oriented view. This is quite natural as people have a tendency to focus on
things and then think about how these things relate to each other; people do not tend
to think of properties or relationships and then identify the things that can be associ-
ated using these properties. However, the development of properties is just as impor-
tant as the classes. Using the example of social networks or transport networks, we
can see that the links between the people or transport hubs are as informative as the
descriptions of the people or hubs themselves. In terms of the pragmatics of ontology
authoring, they are developed hand in hand with the classes they help to describe.
Although Merea Maps has begun creating properties such as “is intended for,” “has
use,” and “has part,” it has not seriously thought too much beyond the subproperty
relationship between “is intended for” and “has use.” We have seen in Chapter 9 on
OWL that, compared to RDFS, OWL introduces additional richness to the way we
can describe a property.
10
As Merea Maps proceeds through its ontology, it becomes
more aware of how it can improve the ontology and indeed the underlying data by
developing more sophisticated properties.
