Re: RFC on 'Semantic Web Locationary'

Ian,

Looks neat. Couldn't get it to work in any of the Semantic Web  
browsers though as the RDF documents don't validate (see [1] for  
example).

The main problem seems to be that you use both rdf:about and rdf:ID  
on the same element, which RDF/XML doesn't allow for some reason.  
Instead of

    <foo:Bar rdf:about="urn:..." rdf:ID="ASDF">
        ...

you should write

    <foo:Bar rdf:about="urn:...">
        <owl:sameAs rdf:resource="#ASDF"/>
        ...

Some URIs also have spaces in them, which will break things. E.g. the  
list in http://www.pipian.com/rdf/places/city/Berlin .

Maybe things would also work better if you served RDF/XML as  
application/rdf+xml instead of application/xml (not sure if the RDF  
browsers can deal with the latter).

On the human-readable side, most of the links on the home page don't  
work for me: http://www.pipian.com/rdf/places/Alaska is 404, and  
http://www.pipian.com/rdf/places/subdiv/Georgia brings up an empty  
screen.

Best,
Richard

[1] http://www.w3.org/RDF/Validator/ARPServlet?URI=http%3A%2F% 
2Fwww.pipian.com%2Frdf%2Fplaces%2Fcountry%2FGermany&PARSE=Parse+URI%3A 
+&TRIPLES_AND_GRAPH=PRINT_TRIPLES&FORMAT=PNG_EMBED


On 5 Feb 2007, at 23:30, Pipian wrote:

>
> A first post I know, but now that the code and data for this  
> project is finally somewhat stable I wanted to see what people  
> think of this mash-up application for basic geospatial semantic web  
> work (which is to say that it's all 'there' if not accessible and  
> completely fleshed out).
>
> A beta version of what I've dubbed 'the Semantic Web Locationary'  
> is available at the URL http://www.pipian.com/rdf/places/
>
> The Locationary is designed partially as a unification effort for  
> the purposes of unifying wide-spread semantic content of  
> geopolitical divisions and population centers (particularly  
> countries and first-order administrative subdivisions), rather than  
> geographic entities in general (for the time being.
>
> Its primary original sources of material include the CIA World  
> Factbook, ISO 3166, Debian isocodes package, and the UN/LOCODE  
> database.  Granted, this makes it relatively simple and examples of  
> these separately are all out there (e.g. those linked from http:// 
> www.daml.org/2001/09/countries/webscriptercolor.html and the entire  
> geonames.org web service), though to my knowledge, no one (except  
> perhaps geonames.org in their human-unreadable format) has linked  
> all three concepts (countries, subdivisions, and cities) for easy  
> static cross-reference in both a 'geopolitical ownership' and  
> 'geographical hierarchy' notation, though the latter is admittedly  
> more subjective than objective.
>
> It's pieced together with pre-existing ontologies (though I can't  
> say that one or two would be nicer if they were removed and  
> 'redone' with another ontology) such as WAIL (http://www.eyrie.org/ 
> ~zednenem/2002/wail/) parts of SWEET (http://sweet.jpl.nasa.gov/ 
> ontology/), and a little bit of the 'Core Communications' ontology  
> (http://dbpubs.stanford.edu:8091/diglib/ginf/1999/05/26-core-comm#)  
> for the purposes of offering a semantic interpretation of the web  
> service response (since the data is not static)
>
> Being a web service, I understand the need to differentiate the  
> intended target from the actual response, and thus added rdf:IDs to  
> the documents to differentiate the subject material of the document  
> from the document as subject.  This of course is something of a  
> mixed message that doesn't seem to have an accepted solution as yet  
> (or am I mistaken these days and a consensus has arisen?)
>
> There's some more critiques of the failings I already recognize in  
> the system at the primary website (http://www.pipian.com/rdf/ 
> places/), but otherwise, it should be both rudimentarily human- 
> navigable (for those with browsers with XSLT support) and machine- 
> navigable for any reasonable query (English mostly at this time,  
> with the exception of countries, which have considerably more  
> accurate foreign name data from the Debian iso-codes package).
>
> That being said, I want to see what other people have to say about  
> the service and how well/poorly I mashed up these ontologies (I  
> particularly wonder about SWEET, as I'm a bit unclear how the  
> owl:imports property in those definitions should be properly  
> interpreted)
>
> --
>
> Ian Jacobi
>
>
>

Received on Tuesday, 6 February 2007 10:27:13 UTC