[Talk-GB] UK address project update

[Rob Nickerson]

Hi all,

I'll address the points raised together:

Firstly, you are both right Tom and Neil. Tom is correct to say that this
is not an import - we are creating a simple editing user interface specific
to addresses and human users will crowdsource the address. Neil is also
correct that it is important to consider conflation. This is important as
we are aiming to make the user interface simple enough that more
contributors can be involved. If we can help avoid any conflict with
existing data in OSM then we make the task easier for newcomers. This is
topic 2 in my September email and I am pleased to say we have an approach
ready. The user interface will look to see if OSM already has an address
within a Land Registry polygon. If it does not then it will show the red
point and ask the user to add the address details, whilst if it does then
the red dot won't be shown and the user can edit the existing OSM address
(e.g. if wrong / missing some details).

And to Christian, hopefully the above makes it clearer that what you
described as "an alternate idea" is precisely where we are heading. The
question is which dots should the user be able to select. As noted in my
email yesterday, we currently think it should be the red dots plus any
address data already in OSM.

Finally, to Neil's second email:

>Would you still expect to add a POI if there is a containing building way?
>What if there is a containing building way with an address?

In the second case, the user interface would show the building with the
address and allow the user to make any changes to the address tags. In the
first case, then it would show the POI ("red point") and add this to OSM as
a node. We are not planning to attempt to merge it to the building way as
the building way might contain multiple addresses (e.g. in the case of
pretty much all homes except detached houses). We want a simple tool and
unfortunately adding capability to split buildings would make it too
complex for many potential new mappers. Of course in the spirit of OSM,
experienced users are free to use a different editor to later split the
building and move the address details from the nodes to the ways. Hopefully
our Cadastral Parcels tile layer will help. See
https://osmuk.org/cadastral-parcels/ .

Thank you,
*Rob*


On Fri, 12 Nov 2021 at 20:46, Rob Nickerson <rob.j.nickerson at gmail.com>
wrote:

> Hi all,
>
> I wrote in September (see [1]) about an address project that OSM UK has
> kicked off. At the time I noted that one of the questions we needed to
> answer was which datasets we should use to get a set of "potential address
> locations" that will then be shown to contributors to help them collect the
> address information. In short there are two options:
>
> A. Open UPRN data
> B. Using Cadastral (Land Registry) INSPIRE parcels to split generalised
> buildings from Ordnance Survey open data, and then taking the centroid of
> the split building.
>
> I am pleased to be able to share a map (
> https://osm-uk-addresses.russss.dev/ ) showing both of these approaches.
>
> The map shows:
>
>    - UPRN points (Green dots)
>    - LR Polygons (Blue areas)
>    - Ordnance Survey split generalised buildings falling within a LR
>    polygon (light brown areas)
>    - The addressable locations from the LR Polygons & OS Buildings
>    approach (Red dots)
>    - The underlying OpenStreetMap map. Note that this means you might see
>    two buildings in some locations: those from OS and those in OSM (grey areas)
>
> Firstly, please note that the red dots for Scotland are currently missing.
> Scotland's cadastral parties are published by ROS and will be added to our
> process early next week.
>
> Secondly, having reviewed a few areas for the Green and Red dots, we
> decided that the Red dots (LR Polygons & OS Buildings approach) gives a
> better set of addressable locations.I use the word "better" very much in a
> relative way. Neither approach is perfect but our view is that the red
> points are a little bit closer to what we want. They have fewer false
> positives which likely matters more. We're keen to get your view - do you
> agree? disagree?
>
> Finally, where the red points fall short are in dense city centres and
> areas with a high concentration of council houses. This is because of a
> lack of cadastral parcels in these areas (e.g. one large cadastral parcel
> for the whole area rather than individual cadastral parcels for each home
> or business unit). We are keen to get your ideas as to how to work around
> this. So far our ideas are:
>
>    1. Identify these locations and direct experienced OSMers to them
>    (leaving the easy places for the new mappers we hope to attract) to add the
>    address by ground survey.
>    2. Or same as 1 but instead of the experienced OSMer having to do a
>    ground survey (as might not live locally) allow them to "pull" the green
>    dots through into the final user interface so that local people new to OSM
>    can collect the address. This could be done by adding these to OSM; that is
>    adding points to OSM that just have a UPRN or similar tag at first so that
>    they can be made available in the map editor we are working on.
>
> Feedback and alternate ideas welcomed. I am particularly keen to know if
> you support idea 2 as typically we have avoided this in the past. For
> example we did not add the postcode centroids to OSM. This time we can add
> an addressable location with a UPRN, city, postcode and/or street name. The
> final details (house name / number) would then need adding via the
> crowdsource project.
>
> I hope this all made sense. If not, I'm happy to answer questions or
> arrange a virtual meeting.
>
> [1]
> https://lists.openstreetmap.org/pipermail/talk-gb/2021-September/027669.html
>
> Thank you,
> *Rob*
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20211113/448fecb4/attachment.htm>