Geography Reference
In-Depth Information
a set of topics listing administrative areas: Frederick Youngs's
Guide to
the Local Administrative Units of England
(1979 and 1991), Melville R ich-
a rd s's
Welsh Administrative and Territorial Units
(1969), and the General
Register Office's
Index of Scotish Place Names rom 1971 Census
(19 75).
18
One oddity revealed by the ADN survey is that of the fifty-two archives
that responded to the relevant question, thirty-two said they were fol-
lowing the NCA rules on place-name indexing rather than an in-house
standard, but the two commonest authorities cited were
Bar tholome w's
Gazeteer of the British Isles
and the Gety Information Institute's
hesau-
rus of Geographical Names,
both of which cover “places” rather than the
administrative units emphasized by the NCA rules.
Although our National Lotery funding required us to create a pub-
lic website with wider popular appeal, a central deliverable was a com-
puterized replacement for the authorities identified by the NCA. This
clearly could not be achieved by modifying our existing ArcGIS system.
First, ArcGIS could not hold a mass of variant names for each entity plus
additional atributes for each name. Second and more fundamentally,
much the largest of the authorities identified by the NCA was Youngs's
Local Administrative Units,
and this contains no maps or coordinates. We
clearly lacked the time to research boundaries for all the units Youngs
lists. Conversely, Youngs provides a mass of textual information on rela-
tionships, especially on hierarchical relationships: district A is in county
B and contains parishes C, D, and so on. We had therefore to develop
an architecture in which hierarchical relationships were required but
boundary polygons were a highly desirable extra and in which entities
could have any number of names.
19
Figure 4.1 uses the parish of Carisbrooke on the Isle of Wight to il-
lustrate the resulting architecture, in particular, the organization of the
system around hierarchical relationships. W hile this structure could be
implemented in any relational database, we actually implemented it in an
object-
relational database, originally Oracle and now Postgres. We could
then use the object extensions to hold the boundary polygons from the
original system as unit atributes. Because the core of this system is a
set of entities rather than a set of terms, and because it holds more than
one kind of relationship, it is not just a thesaurus but an ontology. Our
current production system defines 79,266 units, 129,695 names for those
