[Imports] Belgium address import

Kurt Roeckx kurt at roeckx.be
Sun Nov 24 01:55:03 UTC 2013


On Sat, Nov 23, 2013 at 08:39:20PM -0500, Jason Remillard wrote:
> Hi Ben,
> 
> > - And last but not least an import-plan:
> > http://wiki.openstreetmap.org/wiki/AGIV_CRAB_Import
> 
> I did not see any handling of abbreviations in your python code. If
> the source data has abbreviations, they should get expanded by the
> conversion scripts. Regardless, you should talk about abbreviations in
> the import plan, since they are often issues when doing address
> imports.

This is a good point.  I've already seen cases where we have the
full name currently in osm but the data we get contains some
abbreviatiated form.

I'll try to look at it.  I currently find 970 street names with
a '.' in it.

> Some recommendations.
> 
> - Consider doing some conflation ahead of time in the script. It would
> be possible to eliminate any address that already exist inside of OSM.

I'm already looking at that.

I would also want to add a tag to make a relation with between the
2.  I think this is going to make things easier.

> - You could also make a "conflict" output file, that has identical
> addresses between OSM and the source data and but with positions that
> are above some threshold. Another possible conflict you could spot
> ahead of time, would be seeing if a street name in the address,
> actually had a street with the same name nearby. These will be much
> more work than the non overlapping addresses, having them separate
> insures that they get looked at carefully.

I'm also looking at this.


Kurt




More information about the Imports mailing list