Also, there is Geonames (http://www.geonames.org), which is the primary 
geographic data set on the Semantic Web. Here is the link to Athens:

http://www.geonames.org/search.html?q=athens&country=GR

kc

On 4/6/12 4:54 PM, Karen Miller wrote:
> Ethan, have you considered Getty's Thesaurus of Geographic Names?  It does provide a geographic hierarchy, although the data for Athens they provide isn't quite the one you've described:
>
> http://www.getty.edu/vow/TGNHierarchy?find=athens&place=&nation=&prev_page=1&english=Y&subjectid=7001393
>
> This vocabulary is available in XML here:
>
> http://www.getty.edu/research/tools/vocabularies/obtain/index.html
>
> I have looked at it but not used it; it's a big tangled mess of XML.
>
> MODS mimics a hierarchy (the subject/hierarchicalGeographic element has these children: continent, country, province, region, state, territory, county, city, island, area, extraterrestrialArea, citySection). The VRA Core location element provides a similar mapping.
>
> I try to stay away from Dublin Core, but I did venture onto the DC Terms page just now and saw TGN listed in the vocabulary encoding schemes there, so probably someone has implemented it.
>
> Karen
>
>
> Karen D. Miller
> Monographic/Digital Projects Cataloger
> Bibliographic Services Dept.
> Northwestern University Library
> Evanston, IL
> [log in to unmask]
> 847-467-3462
>
>
>
>
> -----Original Message-----
> From: Code for Libraries [mailto:[log in to unmask]] On Behalf Of Ethan Gruber
> Sent: Thursday, April 05, 2012 12:49 PM
> To: [log in to unmask]
> Subject: [CODE4LIB] Representing geographic hiearchy in linked data
>
> Hi all,
>
> I have a dilemma that needs to be sorted out.  I'm looking for an ontology that can describe geographic hierarchy, and hopefully someone on the list has experience with this.  For example, if I have an RDF record that describes Athens, I want to point Athens to Attica, and Attica to Greece, and so on.  The current proposal is to use dcterms:partOf, but the problem with this is that our records will also use dcterms:partOf to describe a completely different type of relational concept, and it will be almost impossible for scripts to recognize the difference between these two uses of the same DC term.
>
> Thanks,
> Ethan

-- 
Karen Coyle
[log in to unmask] http://kcoyle.net
ph: 1-510-540-7596
m: 1-510-435-8234
skype: kcoylenet