Add key to link third-party website with pictures, menus, description or reviews about POIs

General talk Tagging general discussion
1

I saw the option to adding rating to OSM, like discussed in Ratings for POI's? - OSM Help
I am not thinking about adding rating or reviews directly and I agree with and understand the guideline “do not add subjective information to OSM”

I am thinking about adding link to third-party descriptions/rating/reviews database/websites
(like discussed in [Tagging] contact:* for review websites)
i.e.

amenity=restaurant
reviews:<platform1>=<url1>
website=<url2>

where url1 has to match platforms domain names or be without platforms domain names
in order to limit the scope of what can be included in this tag.

My first guess about the platform to use was:

which was already tightly integrated with OSM and some clients (like OSMAnd)
but apparently is now marked as discontinued (June 2023).

I saw there ware some mention about adding links in the contact:* namespace to “multi-purpose” big online platforms which are considerably inconvenient for multiple reasons:

Maybe a better option would be companies more specifically dedicate to provide information about POI like the usually mentioned three which have been already being referenced in the namespace contact:*:

Regarding POIs offering food, I recently find out about:

which is quite easily accessible and has decent amount of information (at least for Europe), it also has prices for single items in menù sometime.

There are also more open and more general options (but also with fewer contents)

Is the main argument against this key in the line of the following statement?

… adding a lot of tag
containing the pages that talk about this poi makes no sense
and look like seo spam.

I think this can be mitigated by recommending the use of a curated list of such review platforms,
can this data belong to OSM?

NB: not sure if I found all place this topic was already discussed in other threads/pages

2

First, I’d want to thank you for doing relevant research!

It is one of them, but there are more. Problem with ratings in OSM:

  • generally ratings are very subjective, and not on-the-ground verifiable data, so do not belong in OSM
  • by extension, settling on which review site to use, it also subjective/not verifiable
  • those review sites are mostly unrelated commercial entities. Giving priority/exposure to some and not others is dubious at least. Adding all of them is obviously bad idea, and adding just a popular dozen (or two) of them does not really solve the issue.

In short, I don’t think tags should be added in OSM for linking to foursquare, mangroove, tripadvisor and other review sites. If anything, it would make much more sense that they should link from their databases to OSM (and not other way around).

But from user perspective: if you like Tripadvisor, you use that to find places you want. If you prefer Mangroove, use that. Etc.

What however can be done to link that data to OSM from your side, is it that you can create Wikidata entry for that restaurant/amenity (perhaps with just a basic info like type and name; although more if of course preferred, like a picture), and add wikidata=* tag for it to OSM element.

Then all the related info about that entity can be added to that wikidata entry to improve it (e.g. P3134 for linking Tripadvisor ID, or P1968 for Foursquare, or P3108 for Yelp etc.) - which is the appropriate place for adding zillions of related data about some entity (including pictures, menus, owners, history etc.) as little or as much as one wants.

6 Likes
3

I agree on everything that you wrote, but for the sake of completeness,
I reply just to try exhaust all doubts and ramifications and to gather information on this topic for future readers.

The following are question on which I still have doubts:

  1. Should the following data pairs exist in some open database: OSM_id<->(GoogleMaps_id, Facebook_id, Tripadvisor_id, Yelp_id, mangroove_review_id … and similar major national or international services) for OSM POI categories like amenity=* and shop=*?
  1. If yes to 1, is Wikidata the right place to host this data?
  1. What is the best place to store the mentioned pairs between Wikidata and OSM?
  • I am not completely sure why adding dozen of those ids is obviously a bad idea for OSM but is not for Wikidata which has properties for much more than twelve
  • Summary of the current usage of those keys/properties between Wikidata (query, results) and OSM (taginfo summary , sophox (outdated) (using sophox so outdated, now ~2x 4x))
wikidata

image

osm (taginfo)

image

  1. Ideally would be great to have a place where to upload pictures relative to POIs, I am not sure if Wikidata is the right place for a fine grade level of detail similar to OSM, but I guess neither OSM :thinking:

  2. There exists already other standalone open database somehow coupled with OSM? Are them the best option this data (description links, pictures, or detailed information in general)

1 Like
4

I don’t think it’s likely that OSM will have really permanent internal IDs in wide adoption, but it’s just my opinion… But, as even that Permanent ID wiki say, wikidata IDs are best thing we currently have for that. Even if some other better form appears in the future, already having existing wikidata IDs would also allow for much easier (i.e. automated) conversion. So, yeah, go with wikidata is my recommendation.

If yes to 1, is Wikidata the right place to host this data?

Why “if yes to 1”? It seem orthogonal to me (providing you are not putting somewhat fluid OSM node/way/relation ID in wikidata, but instead tagging wikidata permanent ID in OSM (which happens to also be what we do currently for many other cases, even when we don’t care about links to review sites etc).

  • I guess that current Wikidata coverage of such POIs is ~100 times less fine graded than OSM; i.e. see number of restaurants in administrative region of Vienna, OSM ~ 2630, Wikidata~30 or worldwide number: OSM~1.3M, Wikidata~7k

Sure. But it is much younger project, and many people using it also use OSM (“information wants to be free” and all that). Also, as everything else, there is no requirement to add those additional information - even if OSM, people sometimes add just amenity=restaurant point without additional useful information which is in OSM tags (like name, website, opening_hours, contact:*, wikimedia_commons, polygon with real geometry instead of a node etc.) - much less wikidata items which require even more effort and record even more details.

So, it is IMHO to be expected that there will be much less nodes with restaurants linking to review sites and menus (no matter which approach is chosen!) than those without those extra details. I would be quite shocked if reverse turned out to be true (at least in next few decades, while humans are still the primary force editing the map and AIs haven’t replaced them).

IMHO, yes. At least that it seems to me what they claim to want to be: Wikidata:Introduction - Wikidata. Anyway, I do not believe few million extra dataitems for restaurants and fast foods from OSM is going to be a problem for them, as they already have a lots of statements (over 1200M referenced statements and rising steadily). But don’t take my word for it, feel free to ask them if they’d like all OSM restaurants being added to wikidata.

Well, it seems to me to fit (at least) “It refers to an instance of a clearly identifiable conceptual or material entity that can be described using serious and publicly available references” criteria, but, as noted before, feel free to ask there if they’d welcome such information.

  1. What is the best place to store the mentioned pairs between Wikidata and OSM?

As noted, there is wikidata=* key in OSM. Other links / IDs (to review sites etc.) should be in wikidata.org

  • I am not completely sure why adding dozen of those ids is obviously a bad idea for OSM but is not for Wikidata which has properties for much more than twelve

OSM maps things that are on the ground. Review site links are (not usually) posted on the doors of the restaurants. Wikidata, on the other hand, AFAICT has no problem with documenting such “virtual” things, and if fact, seems to welcome them.

Wikidata also allows for linking wikimedia commons pictures (although you could already add those directly in OSM wiki via wikimedia_commons=* tag)

I’m not aware of any other such open global database which would duplicate what wikidata does, especially not an one with past history giving faith that it will remain stable and open (instead of being shutdown or sold to corporate entity that would promptly abuse it).

1 Like
5

Thanks for the complete answer! To be honest also from my point of view Wikidata seems the best option to aggregate and connect additional information to OSM POIs. Given its popularity and great community and foundation behind it. The kind of synergy described here looks wonderful when it will keep working out well at scale (i.e. in terms of query freshness and speed and about the possibility to generate and use offline database chunks).

To this I would argue that also wikidata=* tag value is usually not found on the ground, like reviews site or links to other external databases, but I guess the usage that wikipedia=* tag is the way to minimize this kind of “virtual” information in the OSM database.

Which bring up to me another question, should OSM care about defining how much data overlap should be between wikidata and OSM entries or, like I would expect, the individual guidelines of the two platform apply in each of the two respective entries. I am not speaking about the legal status of copying from one platform to the other but to which extend consider Wikidata an “extension” of OSM database.

To be more concrete, let say a wikidata item representing a POI contains a statement for the website of the place and the related OSM POI contains the wikidata=Qnumber tag but not the website. Should the website and all other OSM approved tag (which values are indipendently known by the mapper) be added to OSM even if they are already present in the wikidata item?

1 Like
6

to which extend consider Wikidata an “extension” of OSM database.

To be more concrete, let say a wikidata item representing a POI contains a statement for the website of the place and the related OSM POI contains the wikidata=Qnumber tag but not the website. Should the website and all other OSM approved tag (which values are indipendently known by the mapper) be added to OSM even if they are already present in the wikidata item?

there is no 1:1 relationship between wikidata and OpenStreetMap, these are related concepts but not necessarily the same in all aspects. You cannot safely assume that all wikidata properties apply to the related OpenStreetMap object and neither can you assume that all properties in OpenStreetMap are applying to the wikidata object. Nor can you assume for example for geographic places that their extent is identical in both databases, especially for objects without formal definition of their boundary - if you accept this, it is logical that also properties like population aren’t the same when the area is different (just to give a simple example, in the real world as in OpenStreetMap, population data is usually referring to areas with formally defined boundaries).

7

Well, that mostly depends on the mapper. Note that OSM and wikidata are different projects, with different editors, and different users (although they sometimes overlap). There is some nice synergy as you note; but they are not related otherwise.
Meaning that there are OSM data consumers which do not consult wikidata, and there are wikidata consumers which do not consult OSM, and there are data consumers which use both.

So from that background, in ideal case both of them will contain maximum amount information possible (e.g. website would be present both in OSM and in wikidata). That way both OSM-only and wikidata-only consumers would reap maximum benefits too.

But, how it is actually used would depend on a (human) editor adding those details. Some fill both datasets to the max, some add most tags in OSM and just put extra data in wikidata, and some add only to wikidata (and maybe link OSM to it as an afterthought, and maybe not even that).

There are AFAIK no rules in OSM what to add in wikidata, nor rules in wikidata what to add to OSM (which would be strange anyway if one project could dictate what other unrelated project should do)

To be more concrete, let say a wikidata item representing a POI contains a statement for the website of the place and the related OSM POI contains the wikidata=Qnumber tag but not the website. Should the website and all other OSM approved tag (which values are indipendently known by the mapper) be added to OSM even if they are already present in the wikidata item?

I personally prefer having at least most usable information directly in OSM (so for restaurants e.g. name, website, opening_hours, cuisine, phone, and maybe wikimedia_commons), as using OSM database is my primary use case.

If I add wikidata with additional information (which happens relatively rarely, compared to my OSM edits), I’d provide there at least name (and hopefully that wikimedia_commons) along with all data which does not have related OSM tag (i.e. the reason I created wikidata entry in the first place).

I’d actually love to include all OSM data in wikidata too, but it just so happens that vast majority of times there are more important things to map in OSM, so wikidata gets just a lower priority…

So, in that particular case, if you have time and will, ideally add website to both. If you don’t - you don’t, just do the things you find most beneficial, whatever those might be.

In any case, if you add/edit data, and such data is duplicated, please take care to keep them in sync. For example, sometimes name in wikidata can be different from name in OSM, which can lead to bad and confusing results in data consumers that use both (e.g. Wikidata translations lead to troublesome labels · Issue #1547 · openmaptiles/openmaptiles · GitHub).

2 Likes
8

Thanks for clarifying, I am aware of that there no 1:1 correspondence in general between OSM/Wikidata objects. What I would expect is to have a tighter correspondence for a subset of objects, i.e. POI, or geographically small objects like man made construction, especially if connected by bidirectional linking (OSM id statement in Wikidata and wikidata= tag in OSM). I am not saying perfect 1:1 match of every tag or statements, but at least having one object containing information which is relevant to the other and vice-versa.

9

I am not saying perfect 1:1 match of every tag or statements, but at least having one object containing information which is relevant to the other and vice-versa.

clearly these synergies can occur, but I wouldn’t go so far as to refrain from mapping properties in OpenStreetMap because they are already present in wikidata. I already had cases where linked wp articles (linked from OpenStreetMap) have been removed from WP, while WD is different than WP regarding relevance, it is still an external unrelated project where we cannot control what happens.

2 Likes
10

And they do occasionally get out of sync: to the point where people wrote a tool to detect those so they can be fixed: https://matkoniecz.github.io/OSM-wikipedia-tag-validator-reports/ (despite the page name, it is about both wikipedia and wikidata). That being said, wikidata usually redirect to other when they are renamed.

I would advice against such bidirectional linking. Not only are OSM ids unstable (not only may the disappear, but the same IDs may become completely different things), but in general trying to setup such bidirectional linking is breaking 2NF and is bound to create conflicts / out-of-sync situations (where OSM points to one wikidata but that wikidata points to some third OSM ID which in turn points to some fourth wikidata etc).

Better to just add (mostly permanent) wikidata=* pointer in OSM, without trying to link back from wikidata to fluid OSM IDs.