Proposed Automated Edit: Merging address points with houses / structures below them for less cluttered mapping

While mapping in my area, i noticed several houses with both a building outline. The points contain the address and government id of the property, while the building / house has no address information.

In my editor, i have the ability to automatically merge the 2 so that all of the data on the address point is applied to the structure outline 1:1 and this can be done in mass quickly.

This problem isnt present in many buildings, ive noticed its more of a town for town discrepancy.
I could fix it easily, just posting this for permission & feedback on the issue.
Please respond with any questions you may have.

Both methods are technically correct, it will often be a choice by the local community. I know places in the world (mostly in Europe) where mappers will routinely compare addresses on OSM against official government datasets (where the licence allows); this often requires addresses on OSM to be presented in a certain way.

You did not specify what your area is but, by looking at your latest edits and from some details in the second picture, we are talking about New York State (USA), here.

8 Likes

I have done some cross checking & research of the imported data, all the address points are accurate based on their position of the structure its giving the address for. I am going to move forward with the operation, right now only for a small part of the Merrick & Freeport NY area. Once i get more responses, i will feel more comfortable doing more merges. Once i do it, i will post the changeset # in this forum.

No, you’ve got that round the wrong way;

You first ask about doing an Automatic Edit Automated Edits code of conduct - OpenStreetMap Wiki, then, when you have the Community on your side, try a test, then if no problems, proceed with the full-scale edit.

12 Likes

Ok, thank you for this, i have not done the edit yet. I will wait for more feedback.

if it is US edit then it would be best to make thread at United States - OpenStreetMap Community Forum

hmm, apparently I can move threads - and I moved it (let me know if it was wrong)

please also retitle it so it mentions automated edit in the title and the first post (or make a new thread)

if you want to edit outside USA - make a new thread mentioning planned edit in the title

1 Like

Hi Chris,

Welcome, and thanks for your work! I’m aware of the context in the other thread with the editor you’ve been working on - I think that it’s relevant here.

Directly to your question, it varies and is an unsettled question. There was at least one prior thread, that I can’t find right now, over some disagreement when merging address points to buildings (or the reverse of extracting them), so it’s an area worth getting input on from the New York community.

On the automated edit side, I think it’d help to have more details on your editor. This function is also easily done in each of the publicly available OSM editors, so some details on how your editor will identify matches and how it merges them would help the community evaluate your request. Concerns would be spatial mismatches, loss of tags, how you handle conflicts between points and buildings when present, are you deleting the point when all tags are moved to the building, are you leaving it if it has other tags, etc.

Thanks!

4 Likes

Just some quick questions:
Will you also be merging nodes for businesses under a building?

How will you handle cases where multiple address nodes exist on a single building? Shops in malls, apartment buildings, duplexes, etc.

4 Likes

Sorry for the wait in response, in my editor, each area has a bounding box, and the check is simple really, if one single point is within the bounding box of an area with any building tag, the option to merge is present, however if any building has more than one address point within the area, it is automatically skipped and left as is.

There are also checks & abilities to specify if the building is in a residential area, commercial area, etc. For the sake of making sure tags are transferred over properly and accurately, each point is directly compared with tags being checked 1:1 one by one and use fallbacks incase a tag couldn’t be transferred and notifies the user (me) that a tag couldn’t be transferred. While i didn’t encounter this problem when doing tests within my editor, since luckily the data import for my area (Nassau County, NY) is relatively clean and accurate, I wanted to make sure that i can always get an accurate result with as much manual approval that is needed, but not too much where it becomes not worth it to spend the time.

Off topic i really want to share more about this editor, but i have really gotten a lot of mixed feedback, some people love it, some think its damaging, some people like the new features, some think it gives mappers too much. Its still private & relatively unknown. It does really seem hard to build trust in this community, mabye i just need to give it more time.

Anyway, thanks for taking the time to respond :slight_smile:

@ChrisTheMapDude please, edit thread title to make clear it is proposing a bot edit

1 Like

Question, when you say “bot” its not really a bot script or mass automatic edit, its not a state automated edit, it would just be small amounts at a time in an area that i specifically select. do “bot” and “automated” have the same meaning basically?

No problem at all - we’re all volunteering together :).

This is less feedback on this specific question and more general feedback that might help you navigate the responses you’re getting.

The OpenStreetMap community has a lot of history with largescale edits that have created problems or left a long legacy of map cleanup. Many of the discussions and disagreements in this forum and in other venues are around what is appropriate to do at scale versus manually as a single person. It’s a long-running tension between those who want to optimize the map by getting good data in quickly and those who want a map that is maximally correct and represents the best human curated data.

I see that you’re coming in and trying to ask questions about what’s appropriate and what’s not, which is honestly the biggest hurdle - the communication piece really helps. It sucks to be repeatedly told no without an avenue to how to get the thing you want to help with done. The people who have been telling you to slow down are the ones who have stepped into the roles to protect the map data from edits that do or may create problems (from licensing, correctness, etc), and then those of us in the local community can step in and help get you oriented to what’s possible instead of what’s not possible.

I understand what you’re trying to do, and I think there are likely still some reasonable questions others may have about it, so I’d suggest continuing to wait and respond to questions. Part of the building trust comes with time in the community. I also think the other part will come with the ability to either see the editor or see the source code. The LLM software era is something new the community at large is having to reckon with right now as bespoke editors become more common - previously, people did write code to perform automated edits, but the specificity and scale were different and typically it required that someone had been deeply involved in the community first to understand the details needed to write the code. That’s flipped on its head now, with LLMs being able to offer editors for niche use-cases and the community’s model of evaluating trust through discussion coming after software is already created is harder for everyone to handle. I do think that before you get full trust from the community for automated edits from your editor, you’ll likely need to provide some detailed information. I understand you’re reluctant right now, but it, or some equivalent alternative will be needed for you to start getting different responses.

Lastly, many members of the US community are here, and many are also in the Slack instance hosted by OpenStreetMap US - if you’re looking for a faster feedback loop with other mappers, you might be interested in joining there too.

9 Likes

I assumed that you will not be manually moving tags after review from address nodes to building areas?

But selecting some area and mass moving them based on some rules with an automated tool?

If you planned to do it manually then it is rather a complete waste of time.
If it is automated, then it is an automated edit and confirming that these rules are appropriate is necessary.

And asking LLM to write such tool and make no mistakes is not enough, we need to know specifically what was taken into account. In actual tested code.

(also, local community may prefer addresses on nodes)

Just in general, there are downsides to what you might think is an overall improvement to OSM.
Each OSM-object has a history. By deleting the current address-node and moving the properties over to the building-way you omit the history of that address. So depending on that history it might be no issue at all or might be something to consider.

Another topic might be address-nodes which are at the same time the entrance-node to the building or a POI.

It might be simply placed deliberately at a certain location to make the routers pick the correct road access point.

I’m not quite understand your method. Checking whether a building and a address-node are in the same “area” sounds not like a good approach. Usually you want to make sure the address-node is within the building-area, and that there is no other address in that building-area.

1 Like

As others have already said, other editors can also do what you are proposing. The reason why it hasn’t been done yet is because there is no need or benefit to do so. Mapping addresses as separate nodes is perfectly valid and well supported by data consumers.
Different local communities have different preferences for which style to use. It is usually a good idea to just follow the style that is already in use in your area.

2 Likes

I would advise a note of caution – not all buildings with only one address mapped on them have only one address.

In my area there are lots of buildings where a shop is on the ground floor and a house is on the upper floors. Typically these are mapped as a building outline + a shop node. The residence is usually not mapped at all (though if my area was better mapped we’d have the home’s entrance mapped with an address). It sounds like your edit might move the shop node onto the building itself which would incorrectly imply the whole building is occupied by the shop.

Would your process be able to avoid these traps?

Edit: Updated wording for clarity

Which shouldn’t really be done.

Apologies, it sounds like I was unclear. I meant that the edit itself would make this error, not a different mapper at some later time. I’ve updated my post to hopefully be more clear

Im going to do more work on the detection, but i did think about shops, buildings with more than one address, and the boundary area the building is in, residential, commercial etc. Im going to still give it more thought, Everyone here is right in saying that it isnt a problem, both ways of mapping is correct, it was just a nitpick from me since in an editor, you notice it but on the normal map, it looks just fine.

I really appreciate the responses.