Flickr Geocodr - inspired by Mikel Marons's work
This version uses k-means clustering to geocode places based on peoples public photos!. There are two webservices:

http://geothings.ning.com/Flickr/flickrgeocodr.php
which can output as REST XML or JSON

http://geothings.ning.com/Flickr/flickrgeocodr_cluster.php
to output more detailed information regarding the clusters themselves in REST XML or JSON.

Example application

Documentation for webservice:
URL: http://geothings.ning.com/Flickr/flickrgeocodr.php?place=
You will need to create a query string. The only required key-value pair is "place", and it will use the other values on their default settings.
place = the placename to search for.
format= json|xml (default xml).
searchmode = tags|text (default tags) (text searchs tags, title and description).
sortmode = interestingness-desc|relevance|(or any other values from Flickr API)(default interestingness-desc).
bbox = any min_lon, min_lat, max_lon, max_lat Comma separated, no spaces. Good for limiting to continent / country.(default world -180.0,-90.0,180.0,90.0).
Tweaks
In addition to changing searchmode and sortmode, results may change and improve by changing the following:
numresults = 5 to 500 (default 50) Number of photo results to process from Flickr, more will slow things down, but may increase sample size, and accuracy.
numclusters = 2 to 20 (default 3) Number of clusters to use internally, experiment with this, I find 3 and 4 fine.
xn_auth=no - a Ning tweak to bypass Ning authenitcation, speeds things up.

results:

<item>
<latitude>51.507715576923</latitude>
<longitude>-0.146583</longitude>
<place>london</place>
<error>1</error>
</item>

Cluster information in JSON (flickrgeocodr_cluster.php)
http://geothings.ning.com/Flickr/flickrgeocodr_cluster.php
You can use all the same GET values as above, but there are additional ones:
showpoints = true|false (defaults to false) Returns photos in each cluster (link to photo, and link url)
showclusters = true|false (defaults to true)
format = xml or json (default is xml)
You can find out what the returned json looks like using this application

Returned XML comes in three flavours, Just the largest cluster, all the clusters and all clusters with point information:

<clusters>
<cluster>
<type>cluster</type>
<id>0</id>
<pointcount>26</pointcount>
<latitude>51.507715576923</latitude>
<longitude>-0.146583</longitude>
<place>london</place>
<bbox>51.459134,-0.224028,51.543732,-0.122985</bbox>
</cluster>
<cluster>
<type>cluster</type>
<id>1</id>
<pointcount>6</pointcount>
<latitude>51.506104833333</latitude>
<longitude>-0.018274666666667</longitude>
<place>london</place>
<bbox>51.483895,-0.055709,51.527369,-0.001266</bbox>
</cluster>
<cluster>
<type>cluster</type>
<id>2</id>
<pointcount>18</pointcount>
<latitude>51.506230722222</latitude>
<longitude>-0.098079611111111</longitude>
<place>london</place>
<bbox>51.421085,-0.119519,51.527963,-0.065188</bbox>
</cluster>
<error>1</error>
</clusters>


http://geothings.ning.com/Flickr/flickrgeocodr_2.php?place=london&format=xml&showpoints=true

<clusters>
<cluster>
<type>cluster</type>
<id>0</id>
<pointcount>26</pointcount>
<latitude>51.507715576923</latitude>
<longitude>-0.146583</longitude>
<place>london</place>
<bbox>51.459134,-0.224028,51.543732,-0.122985</bbox>

<points>
<point>
<type>point</type>
<link>http://www.flickr.com/photos/33286810@N00/13317153</link>
<thumnail>
http://static.flickr.com/11/13317153_ca353f3462_t.jpg
</thumnail>
<latitude>51.50679</latitude>
<longitude>-0.142571</longitude>
</point>
<point>
<type>point</type>
<link>
http://www.flickr.com/photos/87677821@N00/147034740
</link>
<thumnail>
http://static.flickr.com/52/147034740_a53ad71e81_t.jpg
</thumnail>
<latitude>51.543732</latitude>
<longitude>-0.152735</longitude>
</point>
.
.
.

Examples

Example application

Example JSON Cluster query application

Simple geocode query: Where is Yorkshire in the world?
http://geothings.ning.com/Flickr/flickrgeocodr.php?place=yorkshire

Longer query: Where is Southamption in the UK?
http://geothings.ning.com/Flickr/flickrgeocodr.php?place=southampton
&bbox=-8.66250,50.06250,2.58750,58.50000&searchmode=text&sortmode=relevance&numresults=100&numclusters=4&xn_auth=no

Tim Waters (Blog)