[postgis-devel] Geocoder and scripts

[Stephen Frost]

Greetings,

* Lee Keel (lee.keel at uai.com) wrote:
> There have been a few chains going around about using postgis to geo-code
> against Tiger data.  I know that there are some scripts available already,
> but I just thought I would post what we have running now and see what people
> thought.  These scripts were originated from the scripts found under Tiger

Great, thanks!

> http://hud.uai.com/hudpls/ViewMap.aspx?FindAddTI=1.  This is doing a live
> geo-coded search as the user types for the entire US (~39 million roads).

Very cool!

> Future enhancements include:
>   *Street Intersection geocoding

This is certainly a big one that I'd like to see done as well.  It's
pretty much on my todo list right now but I'm not sure when I'll get to
it.  I did adjust things such that it'll return something on city,state
or zip-code lookups, based on TIGER data also.  It could be better
though, I'm sure.  I'm actually pretty annoyed at the difficulty of
getting decent zip code information. :/

>   *FuzzyString match (also broadening the search outwards to match at Zip,
> then City, then State)

Right, the fallback to zip/city,state is something I've done, but it's a
direct fallback, not a search broadening.  I like the idea of search
broadening though, I think that's done to some extent already with the
TIGER stuff (it has fuzzy matching in some places...) but I expect it
could be improved.

>   *Continued growth on normalization algorithms

Yeah, this is seriously a problem.  I'm using basically what the
original tiger geocoder had with a few minor adjustments, but there's
certainly room for improvement here I think.

>   *Use of 'LIKE' will be incorporated if the postgres optimizer changes to
> optimize the use of a % at the end of a string.

I thought it already did, actually...  I'm happy enough with soundex()
and regexp matching though...

> If you would like to test these scripts, simply follow the README.html that
> is in the zip which has basic step-by-step instructions on exactly what to
> do.  Please feel free to contact me if you have any
> questions\comments\suggestions on any of this. 

Interesting..  I have to say that it looks quite a bit different than
the tiger geocoder currently on the PostGIS site!  I'm certainly going
to try to look into it more and understand what's going on.  Do you
happen to recall if there was something which wasn't working well with
the original geocoder?  Also, I'm curious what you're using on the
website for doing the mapping, mapserver?  kamap?  Something else?

Lastly, what license is this under?  I don't see any license statement
in the zip file you sent.  The TIGER geocoder is under the GPL but I'd
like to make sure it's understood what the license is before we use or
look to possibly incorporate anything from this into anything.

Perhaps we'll be able to work together to come up with someone that alot
of people can use and improve and maybe get things like geocoding
intersections added!

	Thanks again, getting this kind of feedback is great!

		Stephen
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.osgeo.org/pipermail/postgis-devel/attachments/20070703/9c3b9369/attachment.sig>