[postgis-users] Your opinion on geocoding ?

[Stephen Woodbridge]

Smith Roman wrote:
>   Hi,
> 
> Thanks, for your replies. I will like to ask for your own opinion as 
> regarding geocoding. Considering the limitations of my data. Which 

I'm not sure about what the limitation on your data are.
Do you have house number ranges? Then you can use the method that Regina 
already suggested.

If you do not have house number ranges then you have a bigger problem as 
roads with the same name might exist in adjacent towns, think Main St. 
Or a road might be multiple segments with the same name, but divided at 
the intersecting streets, think of 6th Ave and all its cross streets.

> geocoding method do you think will be appropriate ? If possible, if you 
> have links to papers on geocoding i will be grateful. Or do u think 
> geocoding the street names using the center point of the linestrings is 
> the most appropriate method ?

To some extent, your geocoding methodology depends on two factors:

1) the quality and detail in the source data
2) what you want to do with the results like zoom a map or generate a 
route to a house, etc.

Some other things you might want to consider are how to handle common 
abbreviation like Street, St., St, Avenue, Ave., Ave, Av., Av, etc. The 
USPS publishes a list of common and acceptable abbreviation.

You might also want to do some fuzzy search using something like 
soundex, or other phonic keys for words. Also how to deal with simple 
typographical error like "West wood" vs. "Westwood" or other common 
spelling errors.

http://www.urisa.org/files/goldberg.pdf
You might also want to read some of the articles listed in their References.

Also notice on the bottom of this page:
http://postgis.refractions.net/download/
a TIGER GEOCODER, there has been a lot of discussion on this in the past 
in the list archives.

Hope this helps,
   -Stephen Woodbridge
    http://imaptools.com/