[Openstreetmap] street level ontology

[Jo Walsh]

On Wed, Nov 17, 2004 at 01:17:30PM +0000, SteveC wrote:
> So as a least worst solution, I'm guessing applet. But at the very least
> the thing can be app-neutral.

well, mapserver+javascript is nice, but will take a bit of fiddling.
An SVG client interface a-la openmap is also something i've  wanted to
do for a while; Ronan's the SVG expert, though. browser-based works
for me... in terms of 'ground-truthing' with handhelds i fantasise
about enough 802.11 coverage; perhaps GPRS as a fallback.

> As for the TIGER definitions, I got the impression they were evolved
> rather than designed and arn't neccesarily the way to go... is that the
> case?

I think they're a good place to start. TIGER isn't comprehensive for
our needs, no, but evolution into place strikes me as a pro rather
than a con ;) I'm not proposeing wholesale re-use of the TIGER CFCC
scheme, but a simpler local alternative using it as a starting place.

There are the ISO 19000 standards for geographic information. They
cost hundreds of euros apiece to look at, and are not royalty free.
sod that! ISO 19000 is also converging with the Open G* Consortium's
GML standards. I'm not a big fan of the geometric portion of GML -
it's overengineered, and not fully supported even in ESRI tools yet.

It is what the OS use for their 'MasterMap', but doesn't express the
spatial semantics that they need; concentrates on geometry and
topology. This is why the OS research department is looking at hybrid
solutions in RDF and OWL, to describe and share proper feature
ontologies... *sigh*. 

Looking at http://www.opengis.org/docs/99-105r2.pdf, which purports to
be the OGC Abstract Feature Specification, it is not telling me
anything conclusive: it "explores how a GIC can formalize and unfiy
its information theoretic foundations to ensure that information
sharing within its community is straightforward". It talks of
"epistemic interfaces". I shudder. 

It points one to the ISO TC211 General Feature Model. Spare a few
hundred euros for an ISO standard, guvnor? anyone remember Z39.50?
*cough*.

In the world of high-level, standard uppoer ontologies there is also
the cyc geography and spatial vocabularies:
http://www.cyc.com/cycdoc/vocab/geography-vocab.html
http://www.cyc.com/cycdoc/vocab/spatial-vocab.html 
i've seen snippets of these in use in the wild - cf
http://dk.space.frot.org/ - but never seen a real-world
application built with them. again, these describe the semantics of
*space*, rather than the simpler semantics of *maps*.

i sense i am ranting. instinct suggests that we look for the simplest
useful thing, which can be exported at the edges into GML, or
RDFgeom2d, or whatever standard is emerging from the OGC and the
convergence efforts. In the meantime, i like the TIGER model of a
simple per-row, per-line-segment classification code scheme for street
level feature types. I don't want to be Not Invented Here, but i just
don't see anything freely available in the world that does what we
want. The OS research department might be persuadable to release some
of the MasterMap ontology if we present them a simple alternative,
commonsense homegrown lite version.

Perhaps the geowankers would know better, or we might provoke an
interesting new three-day flamewar there... ;)


-jo