close
The Wayback Machine - http://web.archive.org/web/20250905184919/https://www.wikidata.org/wiki/Wikidata:Project_chat
Jump to content

Wikidata:Project chat

Add topic
Shortcuts: WD:PC, WD:CHAT, WD:?
From Wikidata


Correct “named after” qualifier to state the more specific object the name refers to

[edit]

The property named after (P138) can be used to indicate the etymology, e.g. for the title of a creative work.

For a work titled The Bald Eagles, which is about the bird of prey species, I would do this:

named after
Normal rank Bald Eagle
0 references
add reference
add value

However, for a work titled The Raptors, with “raptors” representing one or more specific bird of prey species (bald eagles in this case), I’m not sure what property to use as a qualifier to express that information. So far, I’ve used represents (P1268) (recently renamed from “represents” to “representative for”), which doesn’t really fit because it’s intended for people and organizations:

named after
Normal rank bird of prey
represents Bald Eagle
0 references
add reference
add value

keepright! ler (talk) 10:25, 18 August 2025 (UTC)Reply

Reminder for anyone to please take a look at this question!keepright! ler (talk) 08:59, 20 August 2025 (UTC)Reply
Meanwhile, the renaming of represents (P1268) has been undone (now it’s labeled “represents” again), because it actually turns out that the use of this property should not be limited to people and organizations, as the property talk page discussion shows.
However, I still want an answer on whether my use case is proper or not.
keepright! ler (talk) 09:18, 21 August 2025 (UTC)Reply
Why bird of prey represents Bald Eagle? Henrydat (talk) 16:12, 24 August 2025 (UTC)Reply
A work that is titled The Raptors, but it doesn’t depict any other bird of prey and only discusses bald eagles? In such a case, a statement like this would make sense to me.
A real-world case would be Mystery on the Prairie (Q114476732). In German, the episode is titled Die Vögel der Prärie: “Vögel” is the plural word for bird (Q5113), but the episode focuses on a few specific species of bird (check the named after statements on the linked item).
keepright! ler (talk) 08:50, 25 August 2025 (UTC)Reply
In which case you want to use main subject (P921) for what the item covers, which for your example you've already done. You are putting far too many properties into these TV show items, and there is certainly no need to add add named after properties to cover the etymology of a episode title. A children's TV show episode about Meercats, should have perhaps 5 properties, its position in the series, its air date, and that its about Meercats. Everything else should be at the series level. Adding external ID's if other databases work at episode level is also UK. Wikidata will collapse if everything was covered this way. Vicarage (talk) 12:03, 25 August 2025 (UTC)Reply
Keepright! ler We can see here: Bald Eagle can represents bird of prey, so your usage is incorrect. You can use represented by (P1875) with property constraint solving, depicts (P180) or another appropriate property. Henrydat (talk) 13:07, 25 August 2025 (UTC)Reply
Thanks for assisting me! Now that you stated it, represented by (P1875) indeed makes more sense to me, although I will probably not add such information to TV episodes anymore (cf. my comment below).
keepright! ler (talk) 09:51, 26 August 2025 (UTC)Reply
Thanks for the explanation, @Vicarage. I was having a lot of fun with my passionate “Wild Kratts Wikidata” leisure project so far, but I’d be ready to give it up especially if I were able to store all this data in my own Wikibase or a similarly-powerful knowledge graph database. It would be nice of you to answer the following questions of mine:
  • It seems very complicated (especially on the more restrictive macOS) to locally install MediaWiki with Wikibase: Is it possible to use Wikibase for free in conjunction with e.g. Neocities? Or can Neocities only be used for standard homepage-style websites?
  • Should I use a different RDF database solution? So far, I’ve checked out (but haven’t installed) these two options – which one is more suitable as a local database of leisure interests?
    • Neo4j, installed as Neo4j Desktop
    • LinkedDataHub
      • This one seems to require Docker installation for local setup, which is likely nigh-impossible on my macOS computer.
      • Or I have to register an account at a public online instance (which might be a privacy issue).
  • Are there any RDF (graph) database systems that internally work with SQLite and are therefore really easy to setup locally?
  • Also, I’m in the process of migrating to Trilium Notes, which is powered by SQLite and offers basic database features in the form of Attributes, but isn’t sufficient for recording information Wikidata-style.
God bless you! — keepright! ler (talk) 09:46, 26 August 2025 (UTC)Reply
I've never attempted to set up my own wikibase or equivalent, sorry. Vicarage (talk) 10:17, 26 August 2025 (UTC)Reply
@Keepright! ler, Maybe using wikibase.cloud (or use Miraheze, also have Wikibase extensions avaliable as of writing this comment.)
Anyway gonna create a Wikidata item about you based on the statements you created on your user page... DinhHuy2010 (talk) 11:50, 27 August 2025 (UTC)Reply
You think I’m notable enough for my own Wikidata item? I don’t have my own Wikipedia article, after all … and I’m not sure whether my accounts and activity on platforms like Wikidata (Q2013) and OpenStreetMap (Q936) count as serious and publicly available references.
keepright! ler (talk) 08:42, 30 August 2025 (UTC)Reply
Too late, is at Q136000163
Enjoy! DinhHuy2010 (talk) 09:13, 30 August 2025 (UTC)Reply
Already noticed that, I was just a little surprised there … Thanks!
keepright! ler (talk) 09:17, 30 August 2025 (UTC)Reply

Repurposing of Q1370468

[edit]

In November 2023, User:Astirmays changed the description of architectural reconstruction (Q1370468) from "repair of an existing architectural object" to "replicating an ancient architectural object". This made a nonsense of items about historic structures which use it as a value for significant event (P793) after they were repaired or restored.

The linked Wikipedia article is en:Reconstruction (architecture), which is about the former process, according to its opening paragraph, although it acknowledges the latter use also.

What definition are other languages (on Wikidata and Wikipedia) using? What should be done with this item, and those that use it? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:29, 27 August 2025 (UTC)Reply

Hi User:Pigsonthewing, actually, I worked for a while on Commons on categories such as commons:Category:Reconstruction of prehistoric buildings, commons:Category:Reconstruction of ancient Celtic buildings and their subcategories, which are "replicating an ancient architectural object". I needed an upper category and tried also to categorise by country.
Now it appears that perharps it wasn't correct to understand this item as the one for those prehistoric and so replicas. Category:Replica buildings (Q25072425) may not be the appropriate one either, that would be rather commons:Category:Reconstructions of ancient architecture, which needs a Wikidata item. Astirmays (talk) 16:35, 27 August 2025 (UTC)Reply
Items should never be repurposed, and architectural reconstruction (Q1370468) should be returned to its previous description. A new item, something like "historic architecture replication" should have been created. @Astirmays, is it possible for you to go back and fix this issue? Huntster (t @ c) 21:18, 28 August 2025 (UTC)Reply
✓ Done [1], still, I guess there is a lot to do here on Wikidata and more again on Commons, see this one architectural reconstruction (Q1370468), compared with building restoration (Q42553590), architectural conservation (Q2193950), and on Commons: commons:Category:Reconstruction, commons:Category:Reconstruction of buildings, commons:Category:Reconstruction (architecture), commons:Category:Reconstruction (architecture) by country, commons:Category:Reconstruction of buildings by country... Astirmays (talk) 22:06, 29 August 2025 (UTC)Reply
Not really done. this diff shows the changes from just before your November 2023 edits and now.
I don't speak any of the other languages used in labels and descriptions, and I doubt that you speak all of them. Are you confident that they all now align with the changes you just made? Or are some of them translation of what was present in English before your recent changes?
Is has effect (P1542)reconstructed structure (Q117859862) still correct? Is said to be the same as (P460)rebuilding (Q52385658) Are all of the other statements? Do all of the external IDs still apply?
How many items use this one, in the sense that it had between November 2023 and yesterday? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 10:45, 30 August 2025 (UTC)Reply

Multiple items linked to single categories, redux

[edit]

I am reminded that we never resolved the issue discussed at Wikidata:Project chat/Archive/2025/04#Multiple items linked to single categories, where I wrote:

Some Russian monuments are linked to regional, rather than specific, Commons categories.

e.g. Q106547672 and Q106547578, and another 43 items have a Commons category (P373) value referring to c:Category:Cultural heritage monuments in Kondopozhsky District.

Example query: https://w.wiki/DsUv

Can we undo this?

At the time of writing, the above query finds 38,195 such items. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 21:13, 28 August 2025 (UTC)Reply

It looks like some if not all of these were created by AVSBot (talkcontribslogs). That bot is listed as operated by Olksolo (talkcontribslogs).
@Olksolo: can you be of assistance in your bot's edits? William Graham (talk) 18:57, 29 August 2025 (UTC)Reply
In the mentioned discussion, no consensus was reached. Valid arguments were presented against these changes. The current category assignments do not contradict the logic of the elements and the existing constraints for P373. Olksolo (talk) 07:35, 30 August 2025 (UTC)Reply
As I said, the matter is still unresolved.
There are no valid arguments for misusing the category property like this. It is not done for any other type of subject.
The claim "The current category assignments do not contradict the logic of the elements" is false and the question of constraints is the issue which remains to be resolved. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 10:15, 30 August 2025 (UTC)Reply
One can surely return to this question when the constraints change. -- Alexander (talk) 13:11, 30 August 2025 (UTC)Reply
This is a discussion about the constraints. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 17:55, 30 August 2025 (UTC)Reply
I oppose this change because there is no consensus for it. The existing category assignments in the elements of Russian cultural heritage are used, for example, by Commons mobile app for a proper image categorization during upload. If changed, thousands of newly uploaded images will remain uncategorized.
In my opinion, the best solution will be creating specific Commons categories for individual monuments. As soon as such a category appears in the list of cultural heritage, the bot will automatically update the respective Wikidata element and remove the regional category. -- Alexander (talk) 13:20, 30 August 2025 (UTC)Reply
Where is there consensus to use P373 in this manner?
"used, for example, by Commons mobile app for a proper image categorization during upload" The commons app does not require these categories (or any others) to function. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 17:55, 30 August 2025 (UTC)Reply
Image Oppose use of P373 this way. I think this is a misuse of the property. Any app or process that relies on it should be updated to not rely on it. Wikidata doesn't exist to be a kludge around problems (perceived or otherwise) that can and should be resolved at Commons. William Graham (talk) 13:58, 30 August 2025 (UTC)Reply
Funny, that on Commons they generally say in these situations "go to Wikidata and solve it there". Ymblanter (talk) 19:35, 31 August 2025 (UTC)Reply
Do they? Where? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:37, 31 August 2025 (UTC)Reply
Can someone who is a proponent of this arrangement actually explain what the purpose of using P373 this way? I reread all previous discussion and I didn't see a sufficient explanation. William Graham (talk) 22:57, 3 September 2025 (UTC)Reply
The purpose is automatic categorization of images uploaded via Wikidata-based tools (Commons mobile app). -- Alexander (talk) 17:35, 4 September 2025 (UTC)Reply
@Atsirlin can you elaborate? What are the specific technical reasons and architextures? William Graham (talk) 23:02, 4 September 2025 (UTC)Reply
The technical reason is that Commons mobile app takes the category from P373, but it does not read any other parameters, including the monument ID that would otherwise allow an automatic assignment of the category. One can ask for more details here. -- Alexander (talk) 18:15, 5 September 2025 (UTC)Reply

Selecting relevant properties based on the interests of their issuing institutions

[edit]

For a range of different items (ships, spacecraft, buildings, SF authors, magazines, aircraft models) I want to get a list of relevant entries in external databases. For buildings I can do a instance of (P31) selector on Wikidata property for authority control for cultural heritage (Q18618628), but for people I might do (applicable 'stated in' value (P9073)|issued by (P2378)|maintained by (P126)|Wikidata item of this property (P1629)) / (main subject (P921)|field of work (P101)) / science fiction (Q24925). The P31 is clean, but I'm relying on a pre-defined list of property subsets, and I don't want to create new ones for my interesys and the latter approach, while flexible, is a right muddle of connecting properties to their issuing institutions that might be most relevant to my subject, as after all I don't want every database's mentions of Douglas Adams (Q42).

But we don't seem to have a consistent many-to-one way of connecting a property to its issuer that I can rely on. Can we at least agree on a clean mapping of that? Vicarage (talk) 10:56, 29 August 2025 (UTC)Reply

Make a property proposal for it? ArthurPSmith (talk) 16:43, 29 August 2025 (UTC)Reply
My problem is we have too many properties linking properties to organisations. I don't think we need more! Vicarage (talk) 16:50, 29 August 2025 (UTC)Reply

archive team question

[edit]

there a Archive team property id? 2804:7F0:14:1FC2:1563:6DF1:39BD:D99C 01:13, 30 August 2025 (UTC)Reply

By Archive team, what exactly do you mean? Do you have a link? (I'm guessing archive.org things?) ·addshore· talk to me! 09:55, 30 August 2025 (UTC)Reply
There is a separate group called Archive Team, who have a wiki at https://wiki.archiveteam.org/, which we don't have a property for, but I doubt the OP meant that anyway. Vicarage (talk) 13:14, 30 August 2025 (UTC)Reply
yes, i mean exactly this
because they use the mediawiki engine Trollface 2006ALT (talk) 02:19, 1 September 2025 (UTC)Reply
Until some feels there is a need for a property and writes up a proposal, use described by source (P1343) Archive Team (Q4787261) with qualifier URL (P2699) which is better than making it a reference as you did at Sploder (Q134695476) Vicarage (talk) 02:44, 1 September 2025 (UTC)Reply
ok Trollface 2006ALT (talk) 23:53, 1 September 2025 (UTC)Reply

Add a photo

[edit]

How do I add a police clearance letter to a page for someone who was falsely, accused of a crime? Wiggan001 (talk) 16:25, 30 August 2025 (UTC)Reply

You don't need to add a photo here for that. Just using it as a reference (if it is published elsewhere) should suffice. Ainali (talk) 16:38, 30 August 2025 (UTC)Reply
We also don't record accusations, only convictions, though a repealed conviction is OK. Vicarage (talk) 18:51, 30 August 2025 (UTC)Reply

Translating names

[edit]

Good morning, do I have to translate names of small things(e.g tiny settlements) while putting a label in another language? Cause there is one that I could translate, but I don't know should I. Brickguy276 (talk) 21:03, 30 August 2025 (UTC)Reply

No, you don't have to, but it could be helpful if you did so voluntarily,
We can give a better answer if you give us a specific example. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:02, 31 August 2025 (UTC)Reply

Community Appeal: Q131507227

[edit]

Hi everyone,

I'm here to appeal to do the community appeal of a deleted item Q131507227. Please see the discussion.

The summary is: The entity which the item is about is a notable black musician. Some of his fans or regular users has created multiple items for him here on Wikidata. But his main Item was Q131507227 and it was deleted due to being duplicated all the way. The admin refuses to undelete the main item, they have undeleted and deleted the same item multiple times with different reasons.

So I’m creating this topic to appeal the decision, because there’s a discussion if the entity is notable or not, but he is notable in my opinion. I don’t understand why delete an item just because the figure doesn’t have a Wikipedia article yet, but they have a Wiki commons category. BodyGoo (talk) 08:47, 31 August 2025 (UTC)Reply

It's odd that this discussion comprises your only edits to Wikidata. Have you need editing under another account? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:01, 31 August 2025 (UTC)Reply
There seem to be a lot of new SPAs interested in this topic. See also User talk:Bovlb#Page deleted due to duplicated item and related items. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:26, 1 September 2025 (UTC)Reply

gadget to display wikipedia article summaries on wikidata pages

[edit]

I'm experimenting with the wikipedia summary api endpoint https://en.wikipedia.org/api/rest_v1/page/summary/Lion which is proving very useful to display summaries in my websites, so I only display the wikidata description if there is no suitable wikipedia page. It occurs to me to this could be a useful gadget for WD, as API returns json with an html formatted field that could be dropped into a page here. Such a gadget doesn't seems to exist, and as I've only hacked the gadgets of others, never developed one, is there any similar gadget that takes a WD property, feeds it to a URL and parses the resulting json for display, that I could modify? Vicarage (talk) 11:22, 31 August 2025 (UTC)Reply

Not an answer to your question, but isn't the Preview gadget kind of similar to yours? Ainali (talk) 11:56, 31 August 2025 (UTC)Reply
Ah-ha, tool, WikiBridge is what I want. Vicarage (talk) 14:15, 31 August 2025 (UTC)Reply

"instance of" versus "subclass of"

[edit]

Looking at the Beta amyloid (Q417667) entity, the ChEBI CHEBI:64645 webpage says it "is a" polypeptide (Q3084232) but the OWL file (chebi_lite.owl) gives the below RDF/XML. Is it a instance of (P31) relationship or a subclass of (P279) relationship?


   <owl:Class rdf:about="http://purl.obolibrary.org/obo/CHEBI_64645">
       <rdfs:subClassOf rdf:resource="http://purl.obolibrary.org/obo/CHEBI_15841"/>
       <obo:IAO_0000115 rdf:datatype="http://www.w3.org/2001/XMLSchema#string">A peptide of 36-43 amino acids that is processed from the amyloid precursor protein. Appears to be the main constituent of amyloid plaques (deposits found in the brains of Alzheimer's disease patients).</obo:IAO_0000115>
       <oboInOwl:hasAlternativeId rdf:datatype="http://www.w3.org/2001/XMLSchema#string">CHEBI:53002</oboInOwl:hasAlternativeId>
       <oboInOwl:hasOBONamespace rdf:datatype="http://www.w3.org/2001/XMLSchema#string">chebi_ontology</oboInOwl:hasOBONamespace>
       <oboInOwl:id rdf:datatype="http://www.w3.org/2001/XMLSchema#string">CHEBI:64645</oboInOwl:id>
       <oboInOwl:inSubset rdf:resource="http://purl.obolibrary.org/obo/chebi#3_STAR"/>
       <rdfs:label rdf:datatype="http://www.w3.org/2001/XMLSchema#string">amyloid-beta</rdfs:label>
   </owl:Class>

Thanks. Int21h (talk) 01:30, 1 September 2025 (UTC)Reply

If there is an established convention that's what one uses. I count 13 cases of instance of polypeptide and 7 cases of subclass of polypeptide. It's not clear that one exists. If this is a group of compounds and not an unique one, it seems to me it would be more appropriate to use class instead of instance. P31 is intended for leaf-nodes only. One uses P279 for entities that can be used to classify other things. Infrastruktur (talk) 11:59, 1 September 2025 (UTC)Reply
Thank you! Int21h (talk) 15:31, 1 September 2025 (UTC)Reply
subclass of (P279) is the right choice here. All chemical substances are in themselves classes because they are not concrete objects - the instances are the physical instantiations (molecules or pure substances). instance of (P31) should only be used to link to a higher-order class, as is for example done on polypeptide (Q3084232) itself here. That said there are a lot of inconsistencies in the chemical ontology on Wikidata that need to be eventually sorted out. See Wikidata:WikiProject Chemistry and linked archives for previous discussions about this. ArthurPSmith (talk) 13:57, 1 September 2025 (UTC)Reply
Duh that's where I should've posted this. Thank you for responding here! Int21h (talk) 15:32, 1 September 2025 (UTC)Reply

Bibliography

[edit]

Aloha! Where do we put written books, or research papers? We can't fill the Wikipedia with lists, they'll probably try to kill us. I was looking for a list of Mark Twains writings, nothing. But every singer under sun has a discography here. That's a bit strange. Any ideas? --Hedwig in Washington (talk) 02:22, 1 September 2025 (UTC)Reply

haswbstatement:p50=Q7245 [2] or build a query with the query service. That's the value of linked data, stuff like this can be derived live instead of having a person or bot stuff an item with things. William Graham (talk) 04:22, 1 September 2025 (UTC)Reply
Thanks, but only a few people know how to do this. Why not do the same with music? Strange indeed. --Hedwig in Washington (talk) 05:33, 1 September 2025 (UTC)Reply
What is (not) being done with music? --Matěj Suchánek (talk) 15:44, 1 September 2025 (UTC)Reply
See WD:WikiCite. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 11:20, 1 September 2025 (UTC)Reply
Hello, on d:Talk:Q7245 there can be found some Generic queries for authors. M2k~dewiki (talk) 11:41, 1 September 2025 (UTC)Reply
First, this is not about Mark Twain only. What I mean to say is, why are these two mediums treated so differently? There are tons of discographies, but no bibliography. If we do one in a specific manner, we should do the other as well. IMHO that is. --Hedwig in Washington (talk) 04:50, 2 September 2025 (UTC)Reply
There is also a bibliography: d:Q3256326, similar to d:Q960366 list of works or d:Q829965 discographies. M2k~dewiki (talk) 15:40, 2 September 2025 (UTC)Reply
"why are these two mediums treated so differently?" Because there are more people interested in compiling discographies than bibliographies. The vast majority of Wikidata editors are volunteers, whose work is not directed by anyone or anything other than their own enthusiasms. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 15:50, 2 September 2025 (UTC)Reply
Andy, you're right. I like discographies better, also. M2k~dewiki, thanks for digging! -- --Hedwig in Washington (talk) 00:05, 3 September 2025 (UTC)Reply
@M2k~dewiki 2409:40F0:1031:DDC9:8000:0:0:0 08:50, 3 September 2025 (UTC)Reply

Wikidata weekly summary #695

[edit]

Proposal: Implementing a Model Context Protocol (MCP) server for Wikidata

[edit]

Model Context Protocol (Q133436854) (MCP) is an emerging open standard that lets AI agents and applications connect to tools and data sources in a consistent way. By creating an MCP server for Wikidata, we could:

  • Make Wikidata more accessible: any MCP-aware agent (Claude Desktop, IDE plugins, research tools) could query and reuse Wikidata without custom adapters.
  • Support safer contributions: editing tools could enforce references, summaries, batching, and OAuth scopes, aligning with community norms.
  • Enable better reconciliation: MCP can expose reconciliation endpoints so external datasets can be matched with QIDs in a standardized, agent-friendly manner.
  • Promote interoperability: organizations and researchers could plug their systems into Wikidata as easily as attaching a USB device, lowering the barrier for structured contributions.

A first step could be a read-only MCP server that wraps entity lookups and vetted SPARQL queries. Later, we could add edit tools (e.g. add_statement, set_label) with appropriate safeguards.

This would position Wikidata at the center of a fast-growing ecosystem of AI agents and open data tools, while keeping edits aligned with community practices. 5628785a (talk) 13:55, 1 September 2025 (UTC)Reply

You might want to take a look at https://github.com/zzaebok/mcp-wikidata Bovlb (talk) 13:30, 2 September 2025 (UTC)Reply
Thanks for pointing out mcp-wikidata, that looks like a very relevant starting point. From a quick glance it seems to be an early prototype/wrapper, which raises a few questions we might want to explore here:
  • Scope: Should we aim for a minimal read-only server (entity lookups + selected SPARQL queries) as a first concrete deliverable?
  • Alignment with community norms: How could an MCP layer enforce/reinforce existing editing practices (references, summaries, batching)?
  • Hosting/maintenance: Would it make sense for the WMF or Toolforge to host an “official” endpoint, or would this live as a community tool?
  • Interoperability goals: Beyond AI agents, are there particular research or reconciliation workflows that would immediately benefit?
I’d be interested in hearing if others see MCP as complementary to existing APIs (Action API, REST, SPARQL) or more as a unifying protocol layer for external agents. 5628785a (talk) 14:31, 2 September 2025 (UTC)Reply
I see great potential benefits to providing an MCP service for Wikidata, and I have been putting in some thought along the same lines. (I have some experience of implementing MCP servers in my day job.) I would love to provide an MCP wrapper for SPARQL queries, but I have trouble recommending that anyone build services on top of WDQS right now.
This is absolutely the sort of thing that should be hosted as a ToolForge tool. (Another area I have some small experience with.)
One point: I recommend that you stick with the idea of readonly MCP for now, not only because it will make it easier to implement, but also because you will encounter far less community resistance. Bovlb (talk) 17:09, 2 September 2025 (UTC)Reply
Thanks Bovlb, that’s very helpful context. I agree that starting with a strictly read-only MCP service is the pragmatic path:
  • Implementation feasibility: it lowers the technical barrier and makes it easier to ship a working prototype.
  • Community acceptance: it avoids concerns about bypassing established editing workflows while still demonstrating value.
On the WDQS point, I share the concern about stability. A minimal server that focuses first on entity lookups (labels, descriptions, sitelinks, claims with references) could already unlock quite a lot for MCP-aware agents, without putting further load on WDQS. Later we might explore a curated set of vetted SPARQL queries, but entity-level access seems like the safest first deliverable.
I also support the idea of hosting on Toolforge as a community-maintained service, at least initially. That would let us iterate quickly, validate the usefulness for reconciliation and agent integration workflows, and only later consider whether WMF-level hosting makes sense.
Curious if others see specific use-cases where even a basic read-only MCP endpoint would provide immediate wins (e.g. reconciliation in external datasets, IDE integrations, teaching materials)? That could help us prioritize what the minimal scope should include. 5628785a (talk) 19:41, 2 September 2025 (UTC)Reply
Is all your text here LLM written? Because it seems like it and it's very hard to take seriously. William Graham (talk) 22:36, 2 September 2025 (UTC)Reply
I wrote the content myself; it’s just LLM-formatted (i.e., I used a model to proofread, tidy wording and bullet the points). If that style is off-putting, I’m happy to rewrite it in plainer prose. Substantively, the proposal is still: start with a strictly read-only MCP endpoint (entity lookups first), avoid WDQS dependence initially, and host on Toolforge to iterate with the community. If you see specific issues, please point them out and I’ll address them. Here you will find a recent video by Google that explains the MCP protocol if you want more context: https://www.youtube.com/watch?v=IYpt25GHB1U 5628785a (talk) 23:45, 2 September 2025 (UTC)Reply
We've been working on a Wikidata MCP, you can find more information and a Github repository on the Wikidata:MCP. However, this is still work in progress, so feel free to share your thoughts on the discussion page and suggest potential tools we can include.
The MCP will be specifically for read-only access to Wikidata, as we want to keep edits in the hands of the community. Philippe.saade (talk) 10:59, 3 September 2025 (UTC)Reply
Philippe, thanks a lot for your message—this is a great initiative! I agree it makes sense to start with a read-only MCP service.
That said, I do foresee interesting future use cases for hybrid editing, where humans and LLM-agents work hand in hand. For example:
  • Asking an LLM-agent to check all the sources on an item and suggest references for statements that are currently unsourced.
  • Checking whether information is up to date, then (if the suggested sources look relevant) deciding myself to let the agent update the item.
  • Suggesting potential new statements for an item, or evaluate existing ones wrt the ontology.
  • Reconciling a short description against Wikidata to propose the most likely entity (related to vector search and the reconciliation API).
  • Listing all lexemes connected to an item and pointing out missing or unlinked ones that could be created or linked.
All of this would of course be “human in the loop,” but it seems like it could become both useful and productive.
Do you plan to host the first version of the service on Toolforge? 5628785a (talk) 13:17, 3 September 2025 (UTC)Reply

Qualifier and Reference edits hidden from Recent Changes/Watchlists (of other Wikis)

[edit]
Image
What will cause a Recent Changes injection to other Wikis, from a Wikidata Reference or Qualifier edit.

Hello all, please be aware of this upcoming change:
Edits to qualifiers (properties and values) and references (properties and values) within a Wikidata item statement will not produce entries to other Wikis Recent Changes tables or Watchlist notifications.
Wikidata's own recent changes and watchlists remain unchanged.

Explaining the when, what and where is a little convoluted so please see the attached graphic or see the dedicated Meta page for further clarification and direct questions to the discussion page.

This change will be deployed with this week's new MediaWiki version:
MW-1.45/wmf.17, deployed to all Wikis by 05.09.2025. Thank you. - Danny Benjafield (WMDE) (talk) 14:18, 1 September 2025 (UTC)Reply

Contested merge for the same person

[edit]

I've been privately informed by Buszmail that my action of merging Johnny Alegre (Q135974108) (erstwhile labeled "Buszmail" and described as "Founding Head (emeritus) of Pilipinas Panorama Community"; created last 26 August) into Johnny Alegre (Q6266210) was unwanted as he wants to separate his musician's persona from his Wikimedian persona. I boldly did the merge because both items are about the same instance of (P31)human (Q5) and I don't know of any Wikidata policy or guideline that allows a person to have multiple Wikidata items about themselves with the sole purpose of distinguishing different aspects of themselves. The closest thing that I can find is that there is a separate item for J. K. Rowling (Q34660) and her pseudonym of Robert Galbraith (Q110929251), but the latter is notably not instance of (P31)human (Q5) but rather instance of (P31)pseudonym (Q61002) and instance of (P31)persona (Q1077857). Is my merge incorrect or is it valid?

(Also tagging the other users that have edited the merged item, mainly to add labels, descriptions, and aliases in various languages: JWilz12345, Zirconaut, PhiliptheNumber1, RCJourn, Ralffralff, Wolverène, Woordenbrij2, Davidpar, ᱥᱟᱹᱜᱩᱱ ᱗, Aristorkle, Suyash.dwivedi, Skyshifter, Taravyvan Adijene.)

seav (talk) 15:00, 1 September 2025 (UTC)Reply

If they are the same person, then your merge was correct, regardless of the wishes of the individual to maintain a separate persona — Martin (MSGJ · talk) 15:28, 1 September 2025 (UTC)Reply
I've removed some of the unreferenced Wikimedia-related values from Johnny Alegre (Q6266210) (and it's not great that so much else there has no reference). He's primarily known for being a musician, and we should respect and balance that in the way we are describing him. (COI note: I work with him via Affcom.) Thanks. Mike Peel (talk) 21:05, 1 September 2025 (UTC)Reply
You probably should not be editing this item if you have a conflict of interest — Martin (MSGJ · talk) 07:26, 2 September 2025 (UTC)Reply

Confusion regarding Wild Kratts episode order and numbering

[edit]

For TV series where the episodes are supposed to have a specific order because of their plot, the numbering is obvious.

However, Wild Kratts (Q2875960) is a TV series where the episodes mostly don’t tie into each other. In fact, there are many different ways in which they’re ordered in different places:

  • Release date (aka original air date): This is what Wikipedia and the Wild Kratts Fandom wiki currently use. However, they still have different numberings because, for some episodes like Seahorse Rodeo (Q114526149), sources disagree on the air date!
  • Production code numbers
  • IMDb[link in German, I’m unable to access English version] and TVDB each have their own numberings which significantly differ from both production codes and air-date numbering in some places!

As of 1 September 2025, the order and numbering in Wikidata is as follows:

Season 1
Numbering matches production codes except for episodes 1 to 7: Wikidata numbers the pilot episode Polar Bears Don't Dance (Q114546479) as the 1st episode, but it’s the 7th episode according to its production code and original air date.
Season 2
Numbering matches production codes exactly.
Season 3
Numbering matches production codes exactly.
Season 4
Numbering would match production codes if each of the two-part episodes A Creature Christmas (Q114465359) and Creatures of the Deep Sea (Q114477253) were counted as two episodes instead of just one.
Season 5
Numbering matches production codes except for the two-parters Creepy Creatures! (Q114508685) and Hero's Journey (Q114522540), which are only counted as one episode each and are also swapped (so that “Hero’s Journey” is followed by “Creepy Creatures!”).
Season 6
Numbering mostly matches the one on IMDb (as of 24 May 2025), except for these four episodes: Uh-Oh Ostrich (Q114537655), The Great Creature Tail Fail (Q114526802), Adapto the Coyote (Q114542444), Tardigrade Xtreme (Q114523892).
Season 7
Numbering matches IMDb exactly (as of 24 May 2025).

So, which numbering is the “correct” one for Wikidata?
keepright! ler (talk) 10:25, 2 September 2025 (UTC)Reply

Would you please read this and reply?!
keepright! ler (talk) 09:13, 5 September 2025 (UTC)Reply
@Keepright! ler The people at Wikidata:WikiProject Movies might know the answer to your question. RVA2869 (talk) 10:49, 5 September 2025 (UTC)Reply

Memory of the World wiki challenge

[edit]

There is a wiki challenge running during September in which volunteers on Wikidata or Wikipedia, in any language, can win points. We are looking to improve coverage of the UNESCO Memory of the World International Register, which describes the most important documentary heritage from around the world. I have already imported English, French, and Spanish labels from the UNESCO database; users who translate these into other languages can earn points in the challenge. The Khalili Collections (Q63160499) and UNESCO (Q7809) are providing beautiful printed books as prizes. I hope this interests some of you, MartinPoulter (talk) 11:11, 2 September 2025 (UTC)Reply

Isn’t +1 point for each label added to any of these Wikidata items in a new language. rather … erm, easy to game and somewhat suboptimal for Wikidata? Speaking from someone whose gaming the rules resulted in a total overhaul of one of the more important Wikidata competitions … --Emu (talk) 18:25, 2 September 2025 (UTC)Reply

Most linked item has few Wikipedia articles

[edit]

It may amuse folk to know that the most-linked item, according to Special:MostLinkedPages is "scholarly article (Q13442814): article in an academic publication, usually peer reviewed"—but it only has 14 interwiki links to Wikipedias, and two of those are redirects.

Given its ubiquity, perhaps it would be a good idea for more redirects (or articles!) to be created on other Wikipedias, and added to the item?

(The second most-linked item is Europe PubMed Central (Q5412157), with just eight Wikipedia links; the third: English (Q1860); fourth: human (Q5); fifth: English Wikipedia (Q328); sixth: Wikimedia category (Q4167836); seventh: Crossref (Q5188229).) Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 14:09, 3 September 2025 (UTC)Reply

I now see this "Special page" was last updated in September 2019. Is something broken? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 14:15, 3 September 2025 (UTC)Reply
I'm not sure, but it may have been deliberate in response to phab:T213300. —Justin (koavf)TCM 20:46, 3 September 2025 (UTC)Reply
Could also imagine the query is too extensive to run nowadays compared to the effort it gives. Sjoerd de Bruin (talk) 11:23, 4 September 2025 (UTC)Reply
Definitely. The database table is so large I cannot even get the number of its rows (and imagine aggregating them into 118M bins). Related discussion in phab:T245818. --Matěj Suchánek (talk) 16:31, 4 September 2025 (UTC)Reply

Issues creating / New account, can't add lables

[edit]

I cannot create any new items. The interface shows a blank page with no options to add labels, statements, or descriptions.” Peachanda.DuBose (talk) 07:34, 4 September 2025 (UTC)Reply

can you remove any restrictions on my account so that I can create a new label? Peachanda.DuBose (talk) 07:38, 4 September 2025 (UTC)Reply

I cannot create any new items. The interface shows a blank page with no options to add labels, statements, or descriptions.” Peachanda.DuBose (talk) 07:35, 4 September 2025 (UTC)Reply

Editing on mobile is work in progress. Currently you can go to the end of the page and click on where it says "Desktop". This gives you the desktop version of Wikidata which is not very mobile-friendly, but it will let you enter data. At some point in the future this will be improved [3]. Infrastruktur (talk) 09:53, 4 September 2025 (UTC)Reply
Editing of labels and descriptions is possible on mobile. The user never mentions mobile interface. It's something else but we are not given enough information to help. --Matěj Suchánek (talk) 16:21, 4 September 2025 (UTC)Reply
Their edits were tagged as mobile web edits. Infrastruktur (talk) 16:58, 4 September 2025 (UTC)Reply
yes, ok working from my mobile, and even when I try to use the desktop version from my phone, I'm still not getting any options to add lable or anything. It's a black page, basically. 12.187.43.62 23:59, 4 September 2025 (UTC)Reply
It's possible the item is protected from editing by anonymous and new users. Unfortunately, we still haven't got the information which item and how you want to modify it. To create an item, you should be able to use Special:NewItem. --Matěj Suchánek (talk) 06:56, 5 September 2025 (UTC)Reply
[edit]

The Japanese Wikipedia article "インコ" is currently linked on Wikidata to the English article "Psittacidae" (Q7256090). However, "インコ" corresponds more precisely to "True parrot" in English. Therefore, I would like it to be linked to the English article "True parrot". Additionally, the English article "Psittacidae" should instead be linked to the Japanese article "ヨウム科" , not to "インコ". Could a Wikidata administrator please correct these interwiki (sitelink) links? Thank you very much! Malonic acid (talk) 06:56, 5 September 2025 (UTC)Reply

Merge?

[edit]

Are Q131806587 and Q2890573 about the same thing? If so, can someone merge 'em please? This request was posted here by LeChatiliers Pupper. Thanks, Polygnotus (talk) 12:33, 5 September 2025 (UTC)Reply

Based on the date and location, yes. I merged the items. Samoasambia 12:50, 5 September 2025 (UTC)Reply