[talk-au] Victorian Vicmap Address Import Proposal

Andrew Harvey andrew at alantgeo.com.au
Thu Jun 10 10:55:29 UTC 2021


On Tue, 8 Jun 2021, at 7:33 PM, Ewen Hill wrote:
> Andrew,
>   Thank you for both your initial work and the communication as well as the listening. Can I congratulate you on the lib/toOSM.js for the capitalisation and duplication processing. Very detailed
> 
> A few numpty questions after being late to the party...
>  * Is there a node id on the vicmap address and are we storing this in OSM so we can match and look for missing or deleted ones later?

Good question, I've added a section to the code repo at https://gitlab.com/alantgeo/vicmap2osm/-/tree/master#why-no-vicmap-id, essentially because such an ID makes it confusing for OSM contributors, makes the data feel less organic to OSM, and conflation can still be done without it.

>  * Can we provide a couple of samples of the existingAddressesWithNewTagsOnly export once available

Yes I'll post the full import candidates shortly, I'm still working finalising them at the moment.
 
>  * Could we perform some sampling of suburbs/levels. Perhaps 
>    * Mallacoota (large reserves, complex islands and altered crown/public land ownership)
>    * Meringur or Learmouth (large farming community)
>    * Fitzroy (complex inner city)

Yes, I can sample these out, I'll post them shortly.

>  * I can't see what happens on collision with a totally incorrect address. Is the new node just added where you can't find a matching address. 

At the moment yes if it doesn't find an exact match then the new node will be imported. In the conflation I broke the state up by city block (land surrounded by roads) because where there are no addresses in OSM in that block we can be fairly certain there will be no conflicts. Where there are addresses in the block there is a chance that the address exists but it wasn't an exact match, I haven't decided what I think should happen here.

Treating all the blocks where at least one address already exists in OSM would be too overwhelming to review them all manually (636,532 addresses fall into this category compared to 1,311,095 which very likely aren't already in OSM, compared to 107,194 which are certainly already in OSM).

1,311,095 Vicmap addresses are in a block which has no OSM addresses
13,830 Vicmap addresses are within an OSM address polygon (which may or may not be the same address)
636,532 Vicmap addresses are in a block which has some OSM addresses already but couldn't be matched exactly
107,194 Vicmap addresses are in a block which has some OSM addresses already and there was an exact match found

>    * Is there an exceptions file for the above that can be reviewed. Could map roulette some of these?
>  * How have you gone with best practices globally?

Let me circle back to these points. I have created two sets of MapRoulette challenges (not publicly visible yet) but these are for the Vicmap complex name and building / property name, which we aren't importing automatically instead via MapRoulette inviting the local community if they'd like to review these.

> I am amazed by the amount of code you have developed and documented to do this. The benefits of adding this import will far outweigh any minor local issues. Chapeau!

>From the README I wrote, Vicmap data isn't perfect, OSM data isn't perfect. This import isn't without issues, it aims to get it right for the vast majority of instances, but local mappers will still be left with some burden of correcting or cleaning up edge and corner cases. I'm with you that the benefits outweigh the issues.  

I've done on the ground surveying of addresses and shocked by how many errors I made.



More information about the Talk-au mailing list