Database Reference
In-Depth Information
determine the semantic agreement between the user's requirements and those of the
data. It would also not be necessary to convert any data, as in this idealized world
all data would be held as Linked Data, and a GIS would be capable of importing this
directly. If we assume the analysis itself is performed by GeoSPARQL, then it could
be done in this simple case but almost certainly less efficiently than the GIS, and
even this relatively simple case is more or less at the limit of GeoSPARQL's capa-
bilities. For tasks any more complex, it is unlikely that GeoSPARQL would suffice.
And, this example is not only a fairly simple example of spatial analysis, but also in
many respects a simplification of what would be required to perform a more robust
analysis. In reality, it is more likely that the vegetation would graduate from one area
to the next, and this would require spatial modeling and analysis techniques beyond
the capabilities of GeoSPARQL. Put simply, the Semantic Web is not terribly good
at arithmetical analysis, and without GeoSPARQL, the Semantic Web is even more
limited in this respect. The Semantic Web is much more able to process explicit
relationships than perform arithmetic operations, so if spatial relationships in the
data were presented as explicit associations, then it would be possible to obtain the
correct results. This still means that all the relationships need to be precomputed
and stored, quite unrealistic given the sheer number of possible spatial relationships
that exist between objects; we would in effect require anything that can be located
on Earth to be related to all others. Even when considering the initial task of dataset
discovery, it is unlikely that this process would be completely automatic. An ontol-
ogy cannot by its very nature be a complete description of the data or what the data
represents. Therefore, it is likely that the analyst would still want to intervene to
confirm the semantics. Despite these limitations, the Semantic Web will nonetheless
still make this process significantly easier and less prone to error as the data descrip-
tions would all be explicitly defined in a standard machine-readable language. The
Semantic Web is also good at describing the relationships that data has to other data,
at representing the data in a universally uniform manner, and at enabling inference to
be made about the data. It is not a system designed to perform specialist arithmetical
or statistical computation. Thus, although extensions to the Semantic Web such as
GeoSPARQL enable a limited amount of spatial analysis, in many cases the use of a
GIS as an analysis engine will still be required.
The Semantic Web will not replace GIS; each must be used in the most appropri-
ate manner. GIS is there to perform specialist analysis; the Semantic Web is more
about organizing data for maximum reuse.
4.3.2 t
opoloGical
r
elationShipS
Spatial relationships do not solely exist as implicit geometric associations, and there
are circumstances when the Semantic Web can indeed be used more effectively
to perform analysis. Consider a government that releases a number of different
datasets, one containing educational achievements for each school, another show-
ing areas of social deprivation, and another showing health issues by area. The
school information identifies each school by name and address; the areas of social
