[mapserver-users] Geocoding library

[Paul Ramsey]

There are a couple of important things to remember:

1 - The addresses provided by the user and the addresses in your street
network will never ever match.
2 - Your street network does not have the same attribution rules as
other people's street networks.

The general consequence of (1) is that your geocoder will have lots of
rules to map special cases down to a common set of values. (St, Str,
Street => Street) as well as some probabilistic matching for common
misspellings (Hill St ~= Hills St).

The general consequence of (2) is that in (1) you should make your rules
configurable. Your common value for street may be "Street" but mine
might be "St" so that should be configurable so that I can use your
geocoding software with my road network.

The other common problem is a mismatch in how street name schemas encode
modifiers. For example, "1423 N Smith Rd" versus "1423 Smith Rd N". A
common attribute breakdown is NUMBER|PREFIX|NAME|TYPE|SUFFIX. You then
need to be able to break arbitrary strings into the right categories
before geocoding. It's a detailed problem :) Have fun!

Greg Mader wrote:

> For a class project, I am going to try to build a simple geocoding library,
> using shapelib.  My intent is this library could be used in Mapserver to
> return an XY coordinate from a standard address.
> 
> Would this be useful to others?  Does anybody have any advice or ideas?  It
> will be a simple implementation, and not very sophisticated, but I will
> release the source to other, so we could improve it together.
> 
> Thanks for the help on my Mapserver endeavors,

-- 
      __
     /
     | Paul Ramsey
     | Refractions Research
     | Email: pramsey at refractions.net
     | Phone: (250) 885-0632
     \_