Database Reference
In-Depth Information
It believes this will enable its maximal reuse of its data and maximal use by central
and local government and by the environmental and construction companies that
also use its data.
Habitats are usually very closely associated with land cover, so many of the fea-
tures and land cover classifications applied by Merean Maps are of immediate use to
Merean Nature. Merean Nature has already produced an ontology that describes the
habitat classification system that it uses and so begins the process by trying to match
the land cover ontology provided by Merea Maps with this ontology
18
; this process is
known as ontology alignment. What alignment attempts to do is match classes in one
ontology with corresponding ones in the other ontology. This is similar to the process
for Link Discovery (in particular, vocabulary links) outlined in Section 8.6, with the
added complexity of descriptions being in OWL rather than the simpler RDFS.
For example, both Merea Maps and Merean Nature have a class Woodland, so if it
can be demonstrated that these two classes are describing the same real-world objects,
then we can link them together using the predicate
owl:equivalentClass
. This
predicate is similar to the
owl:sameAs
in that it is to classes what
owl:sameAs
is to individuals—it states that they are exactly the same thing. That is, every state-
ment that has been made about one class is also valid for the other class. And, like
owl:sameAs
, it is very easy to misuse and associate two classes that are not quite
the same thing, so care must be taken.
Ontology alignment is still the subject of much research; a significant amount of
this is directed at trying to automate this process (Choi, Song, and Han, 2006; Noy
and Stuckenschmidt 2005, Bouquet, 2007). It is admitted by most that at present the
best that can be achieved is a semiautomated solution with either a human expert
being required to correct an automated process, or the process being used to assist
the expert by suggesting possible alignments between ontologies.
Merean Nature decides to err on the side of caution and align the ontologies using
its experts and does not attempt to use a semiautomated solution.
Figure 10.10
shows
an extract of the habitat class hierarchy that Merean Nature has constructed, which it
has based on the JNCC Phase 1 Habitat system. The system is similar to the land cover
system devised by Merea Maps, but it is not the same. If we look at Woodland Habitat,
one of the properties it has is that the Woodland Habitat is contained within Woodland.
Every Woodland Habitat is only
contained within Woodland.
Class: WoodlandHabitat
isContainedWithin: some Woodland
and
isContainedWithin: only Woodland
19
As indicated, Merean Maps wants to link its concept of woodland to the woodland
concept in Merea Maps' ontology. This will allow it to infer that woodland habitats
exist wherever Merea Maps has mapped woodland. By examining the full definition
of woodland, as specified by Merea Maps, the Merean Nature domain expert is able
to determine that they are the same classes and can therefore be linked together
20:
Woodland and Woodland [Merea Maps]
are equivalent.
Class: Woodland
EquivalentTo: mm:Woodland
21
