[RouterGeocoder] Re: [Pagc-devel] Comments re. Requirements outlined on OSGeo

[(Orkney)Toru Mori]

Dan,

Thank you very much for referring Japanese postal address system.
As project owner of Japanese geocoder (geocoder.ja is part of PostLBS 
http://www.postlbs.org/ ), I want to make my comments on your post.

Yes, you are partially right. The reality gets much complicated. We have 3 
types of address systems and we use 3 types (at least) of Characters sets 
in writing postal address. We do not use street network in geocoding at 
all. They are totally separated.

There are proprietary/opensource geocoding software and web services in 
Japan. They consist of two processes, text normalization and matching with 
address point data (over 30 millions points). Since inputted address text 
may use different characters set and/or way of abbreviation, all the 
address data will be normalized first. This process plays say over 60% of 
total process. Then matching process starts. This can be even done by 
usual database software. 

My purpose in this post to everybody is not to describe how Japanese 
address system is complicated, but want to tell everybody lives in "non-
Asian entity" that even a smart software cannot always handle geocoding 
and routing at a time. We should not forget that almost 60% of world 
population is in Asia.

Mori


> > Once you move to Asian countries, things get complicated. We know that
> > in Japan house numbers along a street are assigned chronologically
> > rather than spatially, as a result the standard algorithms for  
> > locating
> > an address along a blockface breakdown. What is in PAGC (starting with
> > v0.1.4) that could be the basis for a geocoder in Japan (and perhaps
> > other Asian countries) is the parcel geocoder that is now part of  
> > PAGC.
> > However, whether this would work as a solution is largely  
> > speculation on
> > my part, and could well be wrong.