Volker Schmidt wrote:
"We have in Veneto so many bad-data-import problems[...]"
- While I can fully understand your concerns I can assure you I take
care not to import any bad data into the OSM database.
First
of all, before an import I visit the places I plan to import and
survey some addresses on site as a sample to determine how is the
quality of the data to be imported. So far every commune was OK.
Secondly
– I do not import nonsense data – if I see an address lands obviously in
the middle of nowhere, has no complete housenumber/street/city or is
flaw in some other way - I do NOT import it.
Thirdly
– I upload small changesets in order to be able to check if the
process is done right.
Marcello Perathoner wrote:
"OTOH having a house number, even if it's not at the most precise location, is a big win for people who use car navigators: it makes them find the house."
Martin Koppenhoefer wrote:
"I agree with this, of course a housenumber on the entrance is the best if this is where the number has been assigned to, but having it on another point where it is still valid, or very close to it, is almost as good and much better than not having it at all."
- That is exactly my motivation. Being in that region I was surprised,
that such tourism-oriented communes have in OSM very little or even
NO addresses at all, while there is public available address data. Of course it would be perfect if the data would be precise to very last meter, but one should remember that very often streets in that region are very narrow and the surrounding is densely built-up. In that case sometimes even surveying the addresses on foot with a GPS would probably not yield better results as the GPS-signal could be interfered by the buildings.
Marcello Perathoner wrote:
"It is not perfectly clear how you are going to weed out duplicates?"
- First in the to-be-imported data, duplicates would be found by JOSM validator and are handled as follows:
- if I
can I would survey the places with the duplicates before an import and try
to determine the location of the right address,
- if
the above was not possible, I check OSM compatible street-level
imagery,
- if
none of the above options is available (I was not able to be there
physically and no photos available) I skip both of the duplicated
addresses and do not import them at all.
Also before uploading the import, the existing OSM addresses data would be downloaded first, checked for duplicates versus the to-be-imported data, if there are some, the duplicated addresses would not be imported. That would be done before every single changeset being uploaded in order to operate with up-to-date data (should anyone for example add some new addresses manually in the meantime).
Marcello Perathoner wrote:
"Also, the way I'd go about it, would be to write a python script that
downloads the data, massages it, and then either uploads it to OSM
directly or produces an *.osm file for JOSM. It could even check for
duplicates on overpass."
- Unfortunately I'm not that advanced with scripts ;-( So I simple do these things manually. Given the scale of the data it is still quite good manageable. But of course - I wouldn't try it with addresses in Rome or Venice :)
Best regards
Premislav
W dniu 2024-10-07 09:36:04 użytkownik Martin Koppenhoefer napisał:
sent from a phone > On 6 Oct 2024, at 14:17, Marcello Perathoner wrote: > > I don't see any problem in that. Most addresses in my zone are nodes inside buildings and very few are entrance nodes on the building perimeter or even slightly outside. There's little damage done either way. > > OTOH having a house number, even if it's not at the most precise location, is a big win for people who use car navigators: it makes them find the house.I agree with this, of course a housenumber on the entrance is the best if this is where the number has been assigned to, but having it on another point where it is still valid, or very close to it, is almost as good and much better than not having it at all. _______________________________________________ Talk-it mailing list Talk-it@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-it
_______________________________________________ Talk-it mailing list Talk-it@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk-it
