Over the years I’ve received a wide variety of questions related to the Twitter Location Search resources here. Some of them have been pretty straightforward and easy.
But others involve really getting into the guts of how Twitter search really works (especially the cascade), geocoding and all that. So I am adding this page to handle some of the more “advanced” topics. Keep in mind this stuff here is really mostly for the super geeks who are trying to build stuff. It probably isn’t relevant to most Twitter users or marketing usages.
At it’s root, geocoding is creating relationships between physical addresses (which are arbitrarily assigned by humans) and geographic latitude/longitude coordinates (which are specific and agreed upon locations on earth). This is useful for twitter location searchers because software works much better with the mathematical geocode than it does with arbitrary places.
The challenge of geocoding is that it involves all kinds of tedious data related to those arbitrary address indicators. Luckily for us, in the past few years many governments and organizations have been opening up these data sets for use. So it is now possible, if you love the command line and server muckery, to make your own geocoding tools.
One excellent example of these kinds of tools is the OpenAddresses project. It’s a data set handles the relationships for a large number of addresses.
That’s the data side, you’ll also need the processing side. For that you can run Pelias on Ubuntu (I did mention this was for geeks who like command line right?).
From there you’re on your own, but you can do all sorts of stuff including reverse geocoding (finding addresses from lat/long) etc.
For Twitter location searchers, this sort of thing can become useful if you’re working with a specific set of addresses and locations and want to be able to do the research to set up your searches quickly and/or programmatically.