[Talk-GB] addr:place cleanup process

[Rob Nickerson]

Well said. And indeed the use of addr:place is continuing to fall -> down
to 29,412 cases now. Hard to say for sure, but nominatim QA indicates [1]
that at least 15,000 of these are incorrect uses. Add that to the already
fixed cases (47,000) then we see that of the original 76,479 cases of
addr:place in the UK over 80% were incorrect!

JOSM's preset dialogue seems to be a particular issue. The way it positions
addr:place between street and city leads people to think this is something
like a suburb. I have therefore opened a ticket [2] to request this be
changed.

As noted in my earlier emails the OSM UK address editor uses addr:substreet
for the time being so as to guarantee data quality for those data users
that don't want to have to apply some automated algorithm to try and figure
out if a particular addr:place tag is part of the 80% incorrect cases, or
the 20% good cases.

I've also now updated the envelopes map [3] so that it uses both addr:place
and addr:substreet. If both are the same value it only includes one. If you
are seeing both then it is another good indicator that addr:place is wrong
and should be addr:suburb instead.

P.S. I suspect it is worse than 80% incorrect cases but can use that number
as a suitable indicator. It's big enough!!

[1]
https://nominatim.org/qa/#map=4.64/56.61/1.17&layer=addr_place_and_street
[2] https://josm.openstreetmap.de/ticket/21859#ticket
[3] https://envelopes.osmuk.org

Thanks,
*Rob*


On Mon, 31 Jan 2022 at 23:25, Brian Prangle <bprangle at gmail.com> wrote:

> It always seems impossible until it's done: Nelson Mandela
>
> On Mon, 31 Jan 2022, 21:52 Rob Nickerson, <rob.j.nickerson at gmail.com>
> wrote:
>
>> Hi all,
>>
>> Just a quick update on the clean up of the addr:place tag as we end the
>> month. As it currently stands Taginfo GB is showing that addr:place is used
>> 31,187 times, down from 76,479 in late December. Meanwhile addr:suburb has
>> increased from 123,372 to 170,770 uses.
>>
>> As per my email earlier in the month, I do still feel that it is an
>> uphill battle to clean up the misuses of addr:place and as such we should
>> use addr:substreet when we want to record a child+parentstreet type
>> relationship. This is the advice that has gone into the wiki page albeit
>> noting that it wasn't unanimously supported. The drop from 76k to 31k was
>> relatively easy, but increasingly the clean up process will become one of
>> diminishing returns as some 4,500 uses of addr:place only appear in OSM
>> less than 100 times. As in, we have only touched the tip of the iceberg of
>> the incorrect uses of addr:place so far!
>>
>> https://wiki.openstreetmap.org/wiki/Addresses_in_the_United_Kingdom
>>
>> Thank you,
>> *Rob*
>>
>>
>> On Fri, 14 Jan 2022 at 19:24, Rob Nickerson <rob.j.nickerson at gmail.com>
>> wrote:
>>
>>> Hi all,
>>>
>>> There has not really been any comments on the process steps in JOSM so I
>>> don't know if I should proceed or not. I guess I will wait a bit longer as
>>> it's only been a day so far.
>>>
>>> In the meantime, I was thinking about what other approaches we could use
>>> to help with this clean up effort. I realised that there is the OS Open
>>> Names data set which includes a list of populated places (i.e.
>>> settlements). If these values match an addr:place tag then it is a good
>>> indication that the addr:place tag is wrong and should be addr:suburb or
>>> addr:city instead.
>>>
>>> I have therefore made a spreadsheet of all the addr:place tag values
>>> (downloaded from TagInfo GB) that match an OS Open Names "populatedPlace"
>>> value.
>>>
>>> I've marked a few  rows as fixed in the spreadsheet based on the edits I
>>> made to test the JOSM approach. The spreadsheet is open access so feel free
>>> to use this tracker. It is sorted by highest to lowest use of the
>>> addr:place tag.
>>>
>>>
>>> https://docs.google.com/spreadsheets/d/1BGeIyhlL-UODGOVJK543Iy699guc3C6Z0udFE57tYu0/edit?usp=sharing
>>>
>>> Thanks,
>>> *Rob*
>>>
>>>
>>> On Thu, 13 Jan 2022 at 23:51, Rob Nickerson <rob.j.nickerson at gmail.com>
>>> wrote:
>>>
>>>> Hi all,
>>>>
>>>> As we learnt the other week, the addr:place tag has often been
>>>> incorrectly used here in the UK. The tag addr:place is meant to be used
>>>> when the address does *not* reference a street (tagged with
>>>> addr:street) at all. This is frequently seen in small villages in
>>>> continental Europe which often don't have any street names at all.
>>>>
>>>> In the UK most cases of addr:place and addr:street being used together
>>>> are incorrect and are cases where addr:suburb should have been used instead
>>>> of addr:place. You can browse the Nominatim QA
>>>> <https://nominatim.org/qa/#map=8.06/53.26/-0.64&layer=addr_place_and_street>
>>>> to find examples and help clean these up.
>>>>
>>>> To help with this clean up effort I have tested an approach in JOSM
>>>> that some might consider as captured by the Automated Edits code of
>>>> conduct. I therefore wanted to share this process before proceeding.
>>>>
>>>> Step 1:
>>>> Identify a potentially incorrect use of the addr:place tag. To do this
>>>> I am looking at the list of values used on TagInfo GB and checking if any
>>>> of these are villages/towns. If they are then there is a high chance that
>>>> the addr:place tag is incorrect.
>>>> https://taginfo.openstreetmap.org.uk/keys/addr%3Aplace#values
>>>>
>>>> Step 2:
>>>> Open the data in JOSM using an overpass query. An example query I am
>>>> using to download the data is:
>>>>
>>>> [out:xml][timeout:90];
>>>> {{geocodeArea:England}}->.searchArea;
>>>> (
>>>>   nwr["addr:place"="Eccleston"](area.searchArea);
>>>> );
>>>> (._;>;);
>>>> out meta;
>>>>
>>>> Step 3:
>>>> Inspect the data in JOSM paying particular attention to other addr:*
>>>> tags. To do this I start by doing a Ctrl+F and search for e.g.
>>>> "addr:place"="Eccleston". This selects all the features with this tag. I
>>>> can then see the full range of addr:* tags used on these features. High use
>>>> of addr:street reiterates the incorrect use of addr:place. I can also see
>>>> if the addr:suburb tag has been used on any features -> if it has then I go
>>>> investigate this separately. I also investigate uses of all other addr tags
>>>> except housenumber/name, unit, flats and street.
>>>>
>>>> Step 4
>>>> Change addr:place to add:suburb. I get to this stage quicker if step 3
>>>> does not identify any conflicting addr:* tags (e.g. if the only other addr
>>>> tags are housenumber/name and street). The more addr:* tags there are the
>>>> more I look into individual OSM features before making this change.
>>>>
>>>> As always, if in doubt then I don't proceed.
>>>>
>>>> How do you feel about this sort of approach? With proper time spent
>>>> carefully looking at the other addr tags that have been used, then I
>>>> believe it is a low risk edit.
>>>>
>>>> P.S. Feel free to check out the results of a couple trials I made. e.g.
>>>> https://www.openstreetmap.org/changeset/116122672#map=14/53.6500/-2.7281
>>>> I'll pause on making more as I realise I should get the discussion going
>>>> first.
>>>>
>>>> Thank you,
>>>> *Rob*
>>>>
>>> _______________________________________________
>> Talk-GB mailing list
>> Talk-GB at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/talk-gb
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20220212/ca004d75/attachment-0001.htm>