Database Reference
In-Depth Information
the results of the previous two queries, but this is a little messy and certainly not
clear. A more explicit way to handle this is to introduce a new subproperty of “has
part,” such as “has known part” (or some such) that is used to define membership
where the Feature is definitely known to belong to the locality.
Mereology can therefore be used quite successfully if exact boundaries are not
known. These patterns are not complete solutions, and they cannot be applied in
all cases of uncertain boundaries. In some cases, the number of potential members
may be so large that it makes a mereological solution either completely impractical
or at best very unwieldy. In other cases, it may be simply impossible to reasonably
identify the individual features, the classic case being differing types of vegetation
cover that merge together, such as rough grassland merging into scrub, which in turn
merges into woodland. In both these cases, Merea Maps has little option but to resort
to boundary estimations expressed as overlapping polygons.
10.4.6.3 Working with Insufficient Data
There are some things that are very difficult, if not impossible, to define using OWL.
Consider the following two examples encountered by Merea Maps: The first is a
braided river. Braided rivers are defined as rivers that have at least one stretch that
contains multiple channels.
15
We can immediately state the subclass relationship:
Every Braided River is a kind of River. Class: BraidedRiver
SubClassOf: River
But after that, it becomes difficult. The following statements superficially appear to
meet the second requirement for braidedness:
Every Braided River has part a
Braided Stretch.
Every Braided Stretch has at least
two Channels.
Class: BraidedRiver
SubClassOf: hasPart some
BraidedStretch
Class: BraidedStretch
SubClassOf: hasChannels min
2 Channel
The problem is in the way we interpret “multiple channels.” Strictly, any river
stretch that has more than one channel has multiple channels, hence the previous def-
inition that follows this strict interpretation of multiplicity. And, from a mathemati-
cal viewpoint, and even a strict linguistic viewpoint, who could argue? The problem
is that no one would describe a river stretch with just two channels as braided. The
definition uses “multiple” quite loosely, and this reflects the lack of real definition
of braided rivers, or more to the point that it is difficult to answer the question: How
many channels are required to make it braided? No one has ever precisely defined
it: Not two, probably not three or four either. What about five? But, because no one
has defined the exact number, then the question is impossible to answer. People are
able to recognize braided rivers when they see them, so it is possible to classify
them. It can only be done in a precise mechanistic way with great effort, and that has
