Flickr's Geotagging Let Down By Location Search

Thomas Hawk of Geo-tagging mashup Zoomr does a good job of keeping his objectivity in his review of Flickr's new Geotagging facility.

I just had a play with it myself, and I have to agree with Thomas - the nicest aspect is the way that it's integrated with the Flickr Organiser (I still keep spelling that Organizr...) and the by-now-expected Ajax drag-and-drop wizardry. But the most disappointing aspect of it all is the underlying Yahoo! maps data. It's fine if you want to geotag photos down to city block level in most major US cities, but stray off the beaten path to tag photos of mountaineering treks or rock climbing venues - even world-famous rock climbing venues - and the map detail just isn't there.

To be fair, this isn't just a limitation of Yahoo!'s map data, it's similarly limited in Google Maps aswell - and I guess it's a reflection of the underlying business drivers behind the map data. Constructing an index of the whole globe is a massive undertaking, and it has to be funded somehow. The obvious channel is advertising, but only businesses are willing to pay for advertising, and most businesses tend to be centred around urban areas, therefore it's more important to the map provider that the advertisers are kept happy.

It just seems to take on an added dimension of disappointment when this limitation applies to photos. To be honest, one U.S. city block looks pretty much the same as any other U.S. city block, in the grand scheme of things, and part of the whole joy of Flickr comes from discovering fascinating, beautiful images that you may never otherwise have seen. Almost by its very nature, this is going to involve out-of-the-way places like the Sim Gang Glacier in Pakistan, La Dibona of Les Ecrins in the Alps, or K2, which requires 14 days of hard trekking from the nearest road before you even reach the mountain - often cited as the hardest trek in the world, but surely one of the most beautiful. Try to find any of these places in the Flickr map, and you'll have trouble. Even closer to home, in the rugged mountain landscape of North Wales around Tryfan, the map engine draws a blank.

To take the next step in geotagging, and in mapping as a whole, requires the next step to be taken in the search technology that takes a user-supplied string and works out what the hell you actually meant. This is a mammoth task in itself, considering how many different ways people can refer to a particular langitude and latitude, across languages and character sets, let alone local names vs. standard names. But whoever cracks that problem can look forward to a very bright future indeed, and that's just one reason why Natural Language Processing - clearing away the cobwebs of context and language and working with raw chunks of meaning - is becoming such a hot topic right now. Stay tuned...