[Talk-us] Address improvement through imports?

Anthony osm at inbox.org
Wed Nov 2 19:42:51 GMT 2011


On Wed, Nov 2, 2011 at 3:17 PM, Toby Murray <toby.murray at gmail.com> wrote:
> On Wed, Nov 2, 2011 at 11:14 AM, Anthony <osm at inbox.org> wrote:
>> Which brings me to the conclusion that there's no point in importing
>> TIGER address information.  A geocoder can simply try to find the
>> address in OSM, and fall back to TIGER if the address isn't in OSM.
>> Then, once the lat/lon is obtained (possibly from the external TIGER
>> database), it can simply pass the lat/lon back into OSM for routing
>> purposes (possibly along with a warning that TIGER data was used,
>> which is quite likely to be be out of date and/or inaccurate.
>
> I believe this is exactly what Nominatim currently does, minus the
> "warning" part. So address ranges are already a solved problem. All
> we're talking about here is importing *better* address data so that
> Nominatim doesn't have to fall back to TIGER.

Excellent.  Looking back it seems like I might have been the one who
brought up TIGER (although in a quotation, where it was used as an
example).

I checked out the page, and the source I know of for Florida is
already listed.  It's especially good for single family residences,
but it is parcel-based data, so it is of limited use in locations
where there is more than one address at a parcel (which,
unfortunately, is the case for many of the business locations which
people are likely to be looking for).

It's more exact than TIGER, but I still think it should be imported
manually if at all.  And really you can use the same arguments against
importing it as you can use for importing TIGER.  Nominatim (et al)
could just use the Florida county data as a separate database.
There's only an advantage to integrating it into OSM as is unless
you're going to manually verify it as you import it.



More information about the Talk-us mailing list