[stuartcw]

I used to manage address data in a insurance company in Tokyo. We had about 20% of all Japanese addresses in the database.

It was a challenging job.

Parsing the new addresses, normalising them and matching them to existing addresses was a major effort.

The Ministry of Land, Infrastructure, Transport and Tourism publishes electronic data about historical addresses as huge address changes have been done over time with the creation of new cities and metropolitan areas. That all has to be factored in when trying to normalise a pre change address.

Rural addresses get weird. Sometimes we had to try and find anomalies on a map to check that they were possible addresses.

In the end, the biggest help to find mistakes in the data is to send a postcard to that address. If it gets returned then investigate the reason.

[mholt]

As an engineer who did this for US addresses for about 6 years, my hat goes off to you. salute

[jspash]

There is something deeply satisfying to me when a non-technical solution (sending a postcard) is used to solve what could be seen as a purely technical problem.

But I suppose it depends on which department is in charge of the project. In the places I've worked, if the project is deemed to be IT-related in any way, a "meatspace" solution like this would never see the light of day.

[xanipher]

How is this a purely technical problem? I'd argue it's a purely non-technical problem, especially in the many cases where e.g. the postal worker needs some implicit knowledge to make the right decisions.