[Imports] Best practices for address imports

Serge Wroclawski emacsen at gmail.com
Sun Nov 24 18:08:46 UTC 2013

On Sun, Nov 24, 2013 at 12:51 PM, Martin Raifer <tyr.asd at gmail.com> wrote:
> Randal Hale <rjhale at northrivergeographic.com> wrote:
>> I ask because I will (hopefully) be faced with a comparable problem in
>> 2014
> I'm also involved in a similar address import which is still in an early
> phase. And I see more similar projects upcoming (e.g. Italian housenumbers
> by ISTAT: [1]).
> I think that now would be a good time to discuss some good practices for
> address imports in general, wouldn't?
> Here are some concrete questions we are currently facing:
> * If you had an address dataset that positions the addresses at or near to
> the respective building entrances: Wouldn't it make sense to not conflate
> the address data with building outlines? (Because by conflating the
> usefulness of the data would actually be reduced.)

If you know the exact location of the entrance, then you can use the
entrance tag for the address, but in many cases, when we examine this
data on the ground, we find that the entrances are more "near" than
"at"- which means they're useless for finding the actual entrance.

> * What are the pro/cons of a manual community import versus an automated
> import versus some hybrid form?

Generally automated imports are becoming less accepted by the
community. While they have problems, community imports have the idea
that a user will be doing a review of the data and the area.

This is something that community importers need to emphasize more and
work with the community on making sure happens. But a bad community
import is generally better than a good automated import.

> * What are good tools to help with such an import?

Depends on the type.

> * What are good ways to manage quality assurance after the import?

Can you be more specific?

> * Is there a practicable way to keep the imported data in sync with the
> source data?

Yes, but it will depend on the data upstream, and it will depend on
the importers.

> * Can one produce a feedback loop for data providers?

License-wise, this is sometimes difficult but from look at the work
being done in NYC about buildings and addresses.

- Serge

More information about the Imports mailing list