Hi everyone,
I saw the other day the tool from Mateusz Koniecszny https://www.openstreetmap.org/user/Mateusz%20Konieczny called OSM-wikipedia-tag-validator-reports https://matkoniecz.github.io/OSM-wikipedia-tag-validator-reports/Schweiz%20-%20Suisse%20-%20Svizzera%20-%20Svizra%20(Switzerland%20-%20Szwajcaria).html in the WeeklyOSM newsletter and contributed to fix reported issues a bit during the last two weeks (and I'm not the only one, thanks all, we went from ~280 issues to 50).
1) Do we have local rules for Wikidata/Wikipedia
When I asked Mateusz to update the tool (it's done manually), he took the opportunity to ask if Switzerland have some local specific rules regarding Wikipedia and Wikidata. He especially asks :
1. Do you expect wikipedia tag to present if possible or is linking with just wikidata considered as OK? 2. Do you expect wikidata tag to present if possible or is linking with just wikipedia considered as OK?
Do you have an opinion about that? For me, I'm tempted to say that the second statement is OK, but not the first one.
2) Automation for some "obvious" edits
Some errors reported by his validator (and listed on a special page https://matkoniecz.github.io/OSM-wikipedia-tag-validator-reports/Schweiz%20-%20Suisse%20-%20Svizzera%20-%20Svizra%20(Switzerland%20-%20Szwajcaria)%20-%20obvious.html) are so clear that he proposed to have a bot to to fix them in an automated way. Do you see something that is clearly wrong and that should be edited? Do you have something against that?
Best regards,
Imagoiq
Hello,
Le 13.11.22 à 21:47, Imagoiq a écrit :
- Do we have local rules for Wikidata/Wikipedia
my reply isn't about local rule :)
- Do you expect wikipedia tag to present if possible or is linking with just wikidata considered as OK?
- Do you expect wikidata tag to present if possible or is linking with just wikipedia considered as OK?
osm is made with human readable strings and the first "rule" goes against it. i have no problem with an object having a WP tag without wikidata tag.
however, i think that there is no interest to report a "wikidata without wikipedia" nor a "wikipedia without wikipedia" to a human, it's a non-problem that doesn't deserve to spend time on it in my eyes, the user of the data who would like to use one or the other knows easily to find the other element if it is missing. I have no objection to a mechanical edition that would add the missing tag but I don't see any point in it either, the one who uses the data should add it at that moment to take into account the new cases.
- Automation for some "obvious" edits
Do you see something that is clearly wrong and that should be edited? Do you have something against that?
I'm in favour of automated corrections when possible. However, I haven't looked at the list of what he's proposing, and at the guideline level, if he wants to make a mechanical edition, he might as well announce it while respecting the rules.
Regards, Marc
Nov 14, 2022, 09:09 by marc_marc_irc@hotmail.com:
Hello,
Le 13.11.22 à 21:47, Imagoiq a écrit :
- Do we have local rules for Wikidata/Wikipedia
my reply isn't about local rule :)
- Do you expect wikipedia tag to present if possible or is linking
with just wikidata considered as OK? 2. Do you expect wikidata tag to present if possible or is linking with just wikipedia considered as OK?
osm is made with human readable strings and the first "rule" goes against it. i have no problem with an object having a WP tag without wikidata tag.
however, i think that there is no interest to report a "wikidata without wikipedia" nor a "wikipedia without wikipedia" to a human, it's a non-problem that doesn't deserve to spend time on it in my eyes, the user of the data who would like to use one or the other knows easily to find the other element if it is missing. I have no objection to a mechanical edition that would add the missing tag but I don't see any point in it either, the one who uses the data should add it at that moment to take into account the new cases.
For me main benefits are - of adding wikipedia tag: data becomes human readable - of adding wikidata tag: makes easier to fix wikipedia tag in case where Wikidata editors decided to rename article (moves related to dashes, emdashes, brackets and similar are really liked by some)
See https://matkoniecz.github.io/OSM-wikipedia-tag-validator-reports/Schweiz%20-... that could be fixed with automatic edit rather with manual one thanks to presence of wikidata confirming the target of redirect.
- Automation for some "obvious" edits
Do you see something that is clearly wrong and that should be edited? Do you have something against that?
I'm in favour of automated corrections when possible. However, I haven't looked at the list of what he's proposing, and at the guideline level, if he wants to make a mechanical edition, he might as well announce it while respecting the rules.
I was trying to probe individual user to check whether asking entire community makes sense. I guess that I could phrase my question more clearly.
If that would be discussed already and decided then proposing bot edit would make no sense.
Right now no bot edit is being proposed, and would be clearly proposed. No matter answers in this discussion: I would not consider it as approving of any automated edit.
Hi,
On Mon, Nov 14, 2022 at 10:46:03AM +0100, Mateusz Konieczny wrote:
For me main benefits are
- of adding wikipedia tag: data becomes human readable
- of adding wikidata tag: makes easier to fix wikipedia tag in case where
Wikidata editors decided to rename article (moves related to dashes, emdashes, brackets and similar are really liked by some)
See https://matkoniecz.github.io/OSM-wikipedia-tag-validator-reports/Schweiz%20-... that could be fixed with automatic edit rather with manual one thanks to presence of wikidata confirming the target of redirect.
Not all redirects are bad and should be automatically replaced. Sometimes they are there for a reason.
Example: the hamlet of Hurnen In OSM: https://www.openstreetmap.org/node/3099480616
It has a wikipedia tag of de:Hurnen, which is fine. It can be considered a placeholder for its own page which at the moment still redirects to the page of Eschlikon, which it is part from.
The wikidata tag was derived automatically from the wikipedia tag not taking this situation into account. So it's not a reliable source.
The correct handling here would be to delete the wikidata tag.
NB: the fact that the wikidata page has Hurnen and Wallenwil as alternative names of Eschlikon makes me seriously question the usefuleness of wikidata.
Sarah
14 lis 2022, 11:46 od lonvia@denofr.de:
Hi,
On Mon, Nov 14, 2022 at 10:46:03AM +0100, Mateusz Konieczny wrote:
For me main benefits are
- of adding wikipedia tag: data becomes human readable
- of adding wikidata tag: makes easier to fix wikipedia tag in case where
Wikidata editors decided to rename article (moves related to dashes, emdashes, brackets and similar are really liked by some)
See https://matkoniecz.github.io/OSM-wikipedia-tag-validator-reports/Schweiz%20-... that could be fixed with automatic edit rather with manual one thanks to presence of wikidata confirming the target of redirect.
Not all redirects are bad and should be automatically replaced. Sometimes they are there for a reason.
Example: the hamlet of Hurnen In OSM: https://www.openstreetmap.org/node/3099480616
It has a wikipedia tag of de:Hurnen, which is fine. It can be considered a placeholder for its own page which at the moment still redirects to the page of Eschlikon, which it is part from.
The wikidata tag was derived automatically from the wikipedia tag not taking this situation into account. So it's not a reliable source.
The correct handling here would be to delete the wikidata tag.
NB: the fact that the wikidata page has Hurnen and Wallenwil as alternative names of Eschlikon makes me seriously question the usefuleness of wikidata.
oh, there are several things to unpack here:
- so, I think that this wikidata tag should be remove as clearly invalid
- I think that wikipedia tag is also invalid - we should link articles that actually exist not articles not yet created - with neither redirects nor nonexisting nor deleted titles being a valid target
- bot edit that edit wikidata tag was simply wrong and invalid and mistaken https://www.openstreetmap.org/changeset/52343388 was made 5 years ago but I am still tempted to revert it where possible
Adding wikidata matching redirect target is a bad idea and obviously results in adding incorrect data, as some redirects point to more general articles.
So in this case I would want to remove both wikipedia and wikidata tag.
And check whether Eschlikon is linked from Eschlikon OSM object ( https://www.openstreetmap.org/node/30152663 or https://www.openstreetmap.org/relation/1684519 may match)
And yes, wikidata is hideously unreliable. It is useful as stable id for wikipedia article titles, for storing links between wikipedia articles in different languages and their Wikimedia Commons category.
For everything else it is not really reliable and you run into many problems where trying to use it (there are many reasons for that but quality is markedly lower than for example OSM or Wikipedia)
On Wed, Nov 16, 2022 at 10:02:01AM +0100, Mateusz Konieczny wrote:
Not all redirects are bad and should be automatically replaced. Sometimes they are there for a reason.
Example: the hamlet of Hurnen In OSM: https://www.openstreetmap.org/node/3099480616
It has a wikipedia tag of de:Hurnen, which is fine. It can be considered a placeholder for its own page which at the moment still redirects to the page of Eschlikon, which it is part from.
The wikidata tag was derived automatically from the wikipedia tag not taking this situation into account. So it's not a reliable source.
The correct handling here would be to delete the wikidata tag.
NB: the fact that the wikidata page has Hurnen and Wallenwil as alternative names of Eschlikon makes me seriously question the usefuleness of wikidata.
oh, there are several things to unpack here:
so, I think that this wikidata tag should be remove as clearly invalid
I think that wikipedia tag is also invalid - we should link articles that actually
exist not articles not yet created - with neither redirects nor nonexisting nor deleted titles being a valid target
I've fixed a lot of those when going through https://nominatim.org/qa/#map=3.07/0.00/0.00&layer=same_wikidata
The rule I personally applied was: leave the wikipedia tag if there is a significant amount of information on the page it redirects to. What constitutes a significant amount is a bit of a gray area. If there is a dedicated subsection, then that is clearly significant (and I prefer the redirect over linking to the subsection of a page). There are wiki pages that have the entire history of a group of villages. That may be significant for each village. Just the mention that A is a subpart of B is definitely not significant.
So in the particular case of Hurnen, I'd lean towards deleting the wikipedia tag, too.
But that's not necessarily the general rule. So it's probably worth asking for an opinion on redirects in that case from the wider community.
- bot edit that edit wikidata tag was simply wrong and invalid and mistaken
https://www.openstreetmap.org/changeset/52343388 was made 5 years ago but I am still tempted to revert it where possible
I've been fixing a lot of bad wikidata from this import. Note that Yurik did these edits not only in Switzerland but worldwide. Still, I think that the ship has sailed. A revert would be a bad idea now. Rather create a MapRoulette challange of suspicious tags from that import, so that a human can look at them and fix them.
Sarah
Le 14.11.22 à 10:46, Mateusz Konieczny a écrit :
Nov 14, 2022, 09:09 by marc_marc_irc@hotmail.com:
Hello, Le 13.11.22 à 21:47, Imagoiq a écrit : 1) Do we have local rules for Wikidata/Wikipedia my reply isn't about local rule :) 1. Do you expect wikipedia tag to present if possible or is linking with just wikidata considered as OK? 2. Do you expect wikidata tag to present if possible or is linking with just wikipedia considered as OK? osm is made with human readable strings and the first "rule" goes against it. i have no problem with an object having a WP tag without wikidata tag. however, i think that there is no interest to report a "wikidata without wikipedia" nor a "wikipedia without wikipedia" to a human, it's a non-problem that doesn't deserve to spend time on it in my eyes, the user of the data who would like to use one or the other knows easily to find the other element if it is missing. I have no objection to a mechanical edition that would add the missing tag but I don't see any point in it either, the one who uses the data should add it at that moment to take into account the new cases.For me main benefits are
- of adding wikipedia tag: data becomes human readable
ok
- of adding wikidata tag: makes easier to fix wikipedia tag in case where
Wikidata editors decided to rename article (moves related to dashes, emdashes, brackets and similar are really liked by some)
does the rename sometime lost the redirect from the previous name ? if not, the wikidata is useless to follow/fix a redirect in the WP key
I was trying to probe individual user to check whether asking entire community makes sense.
it make sense :)
looking at our list more closely : - follow wikipedia redirect : ok (indeed redirect may be wrong but that's may need a rule that target the issue more closely)
- (only one) wikipedia key in an outdated form : ok
- without wikipedia tag, with wikidata tag present that provides article : I don't object
-wikidata from wikipedia tag : I don't object
- wikipedia tag in an outdated form for removal : ok
many out wikipedia in an outdated form (wikipedia:de', 'wikipedia:en', 'wikipedia:fr') : *not ok* for ex https://www.openstreetmap.org/relation/1686710 should not get an wikipedia=en: but of course the wikipedia=fr: one. I don't know if the information is available or not for all those objects. it would also be necessary to see how the case of regions with more than one official language is treated (there are cases with Bern in the list)
2 suggestions : - add a table of content with the name of all topic - a "fix in one clic" like with osmose and/or an osmose rule (osmose have already some basic check) would be practical, it would have allowed me to solve directly all the review cases
Regards, Marc -- In memory Roger, thanks for all
-
Imagoiq -
Marc M. -
Mateusz Konieczny -
Sarah Hoffmann