This post is not about Caernarfon

Valery Hronusov and the usual suspects over on Google Earth Community have aggregated a number of Google Earth-centric blogs’ feeds, wrung them through‘s geocoder, and turned them into a folderful of KML network links for you.

It’s a praiseworthy effort, but it also highlights the hit-or-miss nature of attempting to coax geospatial information out of an inchoate, semantically bare mass of text. Take, for example, Ogle Earth‘s current list of entries: Some of them contain multiple references to widely disparate locations, some of them contain no references at all, some of them reference place names not in a geospatial context but in a cultural context (First we take Manhattan?), and some of them contain the geospatial reference implicitly, in a KML link. And then there is room for error: The Gombe Chimp blog entry is placed in Gombe, Nigeria, for example.

In sum, I think that in the automated post-processing of blog content is not nearly as effective as the pre-processing of content — i.e. where blog authors ensure that there are no semantic ambigiuities possible when it comes to georeferencing their posts.

It’s just that blog content management systems have been very slow in giving bloggers the tools to let them do this — where, for example, is even the most rudimentary support for GeoRSS or KML out the box in Movable Type, (Google’s) Blogger, MSN Spaces…? Photo-sharing sites have been much faster off the mark.

Some CMSes are more versatile than others, however. Over the past week or two I’ve been playing with pMachine’s Expression Engine, and boy is it powerful, if not without quirks. Sorry to get into specifics now, but it lets you define custom fields, such as for place names and coordinates, and then lets you use these fields wherever you want in your templates, in a variety of contexts.

The result:, A just soft-launched GeoRSS-savvy website about next year’s International Polar Year (IPY). Expression Engine comes out of the box with conditional tags, so for each post, if there is coordinate data, the GeoRSS feed will have it, while on the web the post links to a Google Map, either using the geocoded place name or the coordinates (with the coordinate pair getting priority, if it’s there.)

The main unsolved challenge is dealing with multiple relevant locations in one post. It’s not something that GeoRSS is really set up to handle, as far as I can see. (Here’s my favorite solution for solving this problem, but it would require some rewiring of HTML code). The current setup, at least, lets the author nominate which single location is the most relevant, if any.

One thought on “This post is not about Caernarfon”

  1. Pretty good conclusion Stefan. I think you hit the nail on the head, the wow factor of this stuff initially is pretty impressive, but the long term outlook is much less. Of course the internet is built on this kind of “just because I can” so I guess why not. Of course many of the blogs included in that KML don’t really have any need to be geoRSS compatible (for example Slashgeo).

Comments are closed.