Re: GeoNames and Spatial Queries

On Wed, Feb 25, 2009 at 3:06 AM, John Goodwin
<John.Goodwin@ordnancesurvey.co.uk> wrote:
> Christopher St John wrote:
>
>> I'm looking for feedback/pointers on best practices for
>> finding objects in the Linked Data cloud given a geographic
>> area of interest.
>
> Currently the only mainstream RDF triplestore I am aware of that does
> efficient spatial indexing is Allegrograph (http://agraph.franz.com/).
> Of course this would involve loading your spatial RDF into a local
> triple store.
>

I'll check out the reference, thanks.


> However, I don't think the problem of spatial indexing over the linked
> data web is going to be solved any time soon. We (at OS) think a lot of
> traction can be gained from developing gazetteers in RDF with
> topological relationships between geographic regions
> (http://www.ordnancesurvey.co.uk/oswebsite/ontology/
> http://os.rkbexplorer.com/) (Dbpedia, geonames and freebase all do
> similar things as well). At least then you can ask queries like "find me
> all pubs in England", "Find me hotels in Hampshire and cities bordering
> Hampshire" without the need for geometries or spatial indexing.
>

There's a missing step in going from lat/long (which I have)
to any piece of linked data from which I can navigate.


> Not sure if that answers your question...
>

You've confirmed I'm not crazy.

My worry as a mashup author is that the web of linked data
has holes in places I didn't expect, and navigating across the
gaps requires traditional integration code. That reduces the
advantage that linked data gains from having a generic
representation and common keys. Of course, there will always
be holes, but geodata seems like a pretty big gap given that
Google maps and widely available geodata were the "killer
mashup components" that set off the web 2.0 mashup craze.

I'd still like to get my mashup done, so I'll do a bit more
research then code something. Assuming it actually works,
I'll write up a summary.

Thanks for the feedback, it helps to know I'm not missing
something obvious.

-cks

-- 
Christopher St. John
http://praxisbridge.com

Received on Thursday, 26 February 2009 15:45:11 UTC