[Talk-at] Nicht geschlossene Polygone korrigieren und fehlende Straßen hinzufügen

[TomTom OSM]

Hi everyone,

A few weeks ago I shared some MapRoulette challenges with you in this thread. Keeping in mind your feedback, our editing team is not working on these challenges. I did want to get back to you to respond in detail to the questions and comments that you had, which we have taken a thorough look at in the meantime. This should give you some more insight into our ways of working in general. It will also help me understand your remarks better, so we can further improve our challenges in the future. Your comments are very useful to us, so thank you again.
This time I am not including a German translation, because my message is already quite long.

> We have a saying that the murderer isn't always the one who shoots. As far as I understand, the Maproulette challenges were created by you (or your Tomtom colleagues). And they are already visible to all. It doesn't matter who takes the challenges. It only matters that it happens, and that existing data get damaged. I already gave you an example. That's why you should remove the challenges immediately. - Friedrich K. Volkmann
You are right. Following your remark, we made these challenges undiscoverable, so users of MapRoulette can no longer find them when they are browsing the platform. However, we found that it's not possible to take the challenges offline completely-people in this thread can access them based on the URL. We are looking into a way to make challenges available in read-only mode, so OSM users can provide feedback about the content without actually making any edits. If anyone has any ideas on how to achieve this, these are very welcome.

> Your Maproulette links are useless because the website requires a login to check it out. - Friedrich K. Volkmann
> Ohne Account sieht man zwar nicht alle Details - grubernd
Actually, no separate account is needed to log in to MapRoulette. You can use your OSM account.

> Don't do that! These checks have been done by OSMI and Keepright for many years. All checks not covered by those tools are most certainly flawed (reporting false positives). - Friedrich K. Volkmann
To detect non-closed polygons, we make use of upgraded logic of Atlas Checks<https://github.com/osmlab/atlas-checks>. If you encounter any false positives in the challenge, don't hesitate to let me know. If someone reports an incorrect case, we analyze it and update the logic we use if needed. Also, while working challenges like this, our editors make use of multiple imagery sources (Maxar, Esri, Bing) to validate changes and detect false positives if any.

> How can you conclude that those roads that only exist in Tomtom are real? I strongly suspect that most of those roads are either just wrong, or that the location is so much off that the matching with the corresponding OSM road failed. Please don't flood OSM with unreliable data. We've had hard times cleaning imported junk data out, and we don't want to do that again. Tomtom data sources might even be the same that were already imported and cleaned in OSM. - Friedrich K. Volkmann
Before sharing the challenge, all the missing roads that we detected by comparing OSM to TomTom's map were validated by making use of imagery (e.g. DigitalGlobe) and (where available) probes.

> I've had terrible experience with other maproulette "challenges" already. Users spoiling data they don't understand, in regions they have never been to. - Friedrich K. Volkmann
> In any case, I was wondering which data apart from these "POI"s your editors are planning to use since there are way better OGD available for Austria than what MapRoulette currently offers as aerial layers (although the clarity beta layer might have a decent selection of them but that didn't work when I tested it). - Stefan Tauner
For polygon as well as missing roads challenges, our editors use Maxar, Esri and Bing imagery to be able to make changes with high confidence. If they do not have enough evidence to make a certain edit, they mark the task as "too hard".

> However, I have found some exceptions that indicate some discrepancy in your data that I cannot explain. For example, I have no idea what's supposed to be missing here: https://maproulette.org/challenge/25105/task/124723029 - Stefan Tauner
Thank you for pointing this out. This is a false positive, that should have been detected as such in the validation phase leading up to the challenge creation.

> Something I noticed about the polygons is leisure=pitch + natural=cliff (and thereby not closed) as far as I can see, this combination is quite common, but not documented - not sure, what to think about it
http://overpass-turbo.eu/?Q=way%0A%20%20%5Bleisure%3Dpitch%5D%5Bnatural%3Dcliff%5D%0A%20%20(%7B%7Bbbox%7D%7D)%3B%0A%20%20(._%3B%3E%3B)%3B%0Aout%3B&C=47.00554;10.345;9
> As for the streets, you should also check for aeroway=taxiway to avoid false positives. - andreas wecer
Thank you for bringing these cases to my attention. We are looking into them and will update our logic for future challenges if needed.

> Maybe you should emphasize a bit more in the instructions, that the solution doesn't necessarily have to be to close some polygon, because the tagging could be wrong in the first place (or at least unusual). While the generated warnings are correct, it might not always be as obvious as with this broken zoo and even there you would have to separate the zoo object from the fence and not just close the fence - andreas wecer
The detailed instructions for mappers in the MapRoulette challenge (see below) are more extended. Are these clear in your opinion?
"Features intended to be a polygon should be correctly tagged. When the tag indicates that the object should be a polygon, but the geometry is built out of a NON-closed way, an error is logged.
The possible causes of such violation include:

  *   Non-area feature has a tag that incorrectly implies an area geometry.
  *   Geometry of a way is not closed.
  *   Geometry of either outer or inner ring in a relation, potentially composed of multiple ways, is not closed.
  *   Relation has a member with incorrect or missing role."

> Similar to the combination I posted earlier, I've found unconventional objects with playground=slide + leisure=playground on single ways for example:
> https://overpass-turbo.eu/s/1g2f
> https://osmcha.org/changesets/117221727
> Contrary to the cliff example, this change from daGreif isn't necessarily wrong, as playground slides can be tagged as polygons and you could argue that the whole playground consists only of this long slide, but I'm not sure if this is really an improvement. An alternative could be attraction=slide, which is used a few times for similar big slides, but isn't documented either:
> https://overpass-turbo.eu/s/1g79 - andreas wecer
We looked at this change and saw that this edit by daGreif was made after we made available our challenge, but not as a part of our challenge.
Regardless, if there are certain rules we should take into account for playground slides, we are happy to do so. Let me know if you have more examples, or if you find any documentation based on which we should update our logic for future challenges.

> Another thing that comes up a few times is to mix up the keys "water" (which should be an area) instead of "waterway" for streams. - andreas wecer
Could you provide any examples of this from the challenge?

In general, if you spot any issue in a challenge of ours, you can always inform me about it and I will have our team look into it. You can also contact me with any other questions.

Marjan


From: TomTom OSM via Talk-at <talk-at at openstreetmap.org>
Sent: Thursday 17 February 2022 8:53
To: OpenStreetMap AT <talk-at at openstreetmap.org>
Cc: TomTom OSM <OSM at tomtom.com>; andreas wecer <andreas.wecer at gmail.com>
Subject: Re: [Talk-at] Nicht geschlossene Polygone korrigieren und fehlende Straßen hinzufügen


Hello everyone,

Thank you again for your thoughtful comments and questions.

We want to make sure that our contributions to OpenStreetMap in Austria are valuable, and that our procedures are transparent. For that reason, we have taken the challenges offline, and we will take the time to process your feedback.

I will reach out to you again when we have more questions, or if we want to share a new project, for which we will keep in mind the valuable input that you have given us.

Marjan

---

post-edited German translation from DeepL Translate - www.DeepL.com/Translator<https://eur02.safelinks.protection.outlook.com/?url=http%3A%2F%2Fwww.deepl.com%2FTranslator&data=04%7C01%7Cosm%40tomtom.com%7Ccf0b5758490442b42e3c08d9f1ead63e%7C374f80267b544a3ab87d328fa26ec10d%7C0%7C0%7C637806814244461754%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=A83JFTWVMhH0JXxsGTQula2qIUq2T5Zjo6p2KfRAIS8%3D&reserved=0> (free version)

Hallo zusammen,

nochmals vielen Dank für Ihre nachdenklichen Kommentare und Fragen.

Wir möchten sicherstellen, dass unsere Beiträge zu OpenStreetMap in Österreich wertvoll und unsere Verfahren transparent sind. Aus diesem Grund haben wir die Challenges offline genommen und werden wir uns die Zeit nehmen, Ihr Feedback zu bearbeiten.

Ich werde mich wieder an Sie wenden, wenn wir weitere Fragen haben oder wenn wir ein neues Projekt vorstellen möchten, für das wir den wertvollen Input, den Sie uns gegeben haben, berücksichtigen werden.

Marjan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-at/attachments/20220225/a12cc00f/attachment.htm>