A lot of people seem to be blogging about MetaCarta recently and I've finally had a spare 10 Min's to take a look at their labs section where you get to play with the technology, and I have to say its pretty neat. The API is simple but functional, and makes it easy to do the easy things, although what its doing is very cool. Its pretty quick geocoding a document, considering the amount of text there can be. You can access the labs and start playing here

MetaCarta Labs
You can pass in a document, a text query or a URL, and it will return either an image location map, GML or Javascript response for the locations. I cant run Javascript on this blog, so the best I can offer is the image below which should update dynamically to show the current geocoding of this blog.

As has been pointed out by others the main issue with this, as with any geo-coder is the data thats used behind the system. I've worked quite a bit with our own Gazetteer toolkit in the UK (Gazops) which is only ever as good as the data that sits inside it, and invariably this data is not cheap, at least in the UK its not. though maybe one-day….

I guess it's not so important when geocoding things on a global scale to place-names, but for many people addresses are key, especially those hard to find ones and non-addressable locations. Its not entirely clear what data MetaCarta are using, it looks like it will do global place-names and addresses in the US. There is this research paper, which describes the approach they have taken and some of the data incorporated into the system. What would be really cool, would be if it was possible to plug a custom gazetteer database behind the NLP part of the system, it may be possible as they seem to do quite a lot of products aimed at the enterprise and can supply custom Data modules such as energy or US address data They also have the inevitable Google Earth plug-in as well as an ArcGIS extension.

Technorati : , ,