Wikipedia talk:Wikidata/2017 State of affairs/Archive 7

Archive 1

←

Archive 5

Archive 6

Archive 7

Archive 8

Archive 9

Archive 10

UNREADABLE WIKIDATA REFS

There are currently 225 transclusions of Template:Cite_Q. In the article-read view it looks like a standard detailed ref. However if you try to search for the ref-text in the wikitext, you can't find it. In the wikitext it looks like this:

{{cite Q|Q29581755}}

which renders as:

Lois Reynolds; Tilli Tansey, eds. (2008). Superbugs and Superdrugs: A History of MRSA. Wellcome Witnesses to Contemporary Medicine. History of Modern Biomedicine Research Group. ISBN 978-0-85484-114-1. OL 23194335M. Wikidata Q29581755.

A relatively new user would have no clue what the frick is going on. And as an experienced editor with some wikidata experience, I still had no frinkin clue what was going on when I couldn't find the ref in the wikitext. I eventually found the cite Q at the ref location.

How many people even know this was going on???

My first thought was to nominate it for deletion discussion. Then I figured it probably warranted a Pump discussion. Which brings me back here, and the messy question of whether to run a narrow RFC on this wikidata issue or whether to wait for a comprehensive wikidata RFC. Alsee (talk) 05:39, 12 September 2017 (UTC)

I do not have a strong opinion either way - it could be all moved to Wikitext, or it could be left like this so that other WMF projects can use it in the same way - since this ref is unlikely to change, there is probably not much added value in sharing it between the projects. I must note however that in complex templates which we have plenty it is very often also impossible to understand where different features come from, and even going to the template code does not help. Several times I wanted to correct info (most often in articles related to railway stations) and I had to give up since I could not figure out how info could have been modified. This is approximately as usable as the Wikidata code.--Ymblanter (talk) 05:54, 12 September 2017 (UTC)

@Mike Peel: who is I believe responsible for creation of most of these citations (they are apparently coming from moving Infoboxes to Wikidata).--Ymblanter (talk) 05:59, 12 September 2017 (UTC)

I happen to think that it is a great idea to consolidate citation metadata on Wikidata, rather than copying it separately into each individual Wikipedia article that cites something. That way, errors, partial information, or new information in a citation would only need to be corrected once, rather than tracking down all the different copies of a citation and correcting each one separately. For instance, when I create biographies on academics it typically takes me some time to find all of the citations of the subject's works on different Wikipedia articles and add authorlinks for them; reducing this to once per publication rather than once per citation of a publication would be a great help. So I would be strongly opposed to ripping all this out and going back to the bad old way of doing things. On the other hand, if we could make these easier to find in some way, that could be helpful. Right now you have to notice the Wikidata Q-number at the end of the citation in the article view. But really, is that so different from the things like <ref name=":-0"/> with cryptic ref-names that we also have scattered throughout our articles?

I do have a technical question about the implementation of these citations, that I couldn't figure out looking at some of these in Wikidata. Are publication authors stored only as Q-numbers for the author identity, or are they also stored as strings describing how the author's name was written on that specific publication? For instance, The example you give has authors "Lois Reynolds; Tilli Tansey, eds." but actually if you look at the publication it refers to they are listed there as "L A Reynolds and E M Tansey". In this case the changed author spelling is relatively harmless but in some other cases where people have made more significant changes to their names (for instance, by marriage), using only the Q-number could lead to significant mangling of citations. I also wonder why the volume in the series isn't included in the article view but that's a minor and probably easily fixable bug; it's there in the wikidata. —David Eppstein (talk) 06:30, 12 September 2017 (UTC)

@David Eppstein:, I do not know the answer to your question (I rarely work with citations on Wikidata), but I am sure the answer is known. Do you want me to ask on Wikidata?--Ymblanter (talk) 10:32, 12 September 2017 (UTC)

Please nominate it for deletion and get rid of it completely. Hardly anyone is watching whether Wikidata info used in our articles gets changed at Wikidata, meaning that this is yet another inroad for unwatched changes. Basically, we are more and more becoming a mirror of Wikidata, but with a disconnect between the mirror (which is what 99% of the readers see and where 99% of the editors are) and the source (where a few people are getting their data from, but where hardly anyone watches for vandalism or other unhelpful changes).

I see that this is for example used on Template:Infobox World Heritage Site, where it creates things like Regensburg: area: [7]. Yep, really helpful! The template has an "edit on Wikidata", but apart from the indication that yes, this is a World Heritage Site, it is not clear at all where this should be edited, or where the source (the [7] is a source) comes from. This is extremely un-editorfriendly and no improvement at all. (In other cases, the "area" is simply the area of the whole city, not of the world heritage site, but this uses Wikidata, not a cite Q template, so is not directly relevant here, except as an indication that relying on Wikidata agian creates errors).

Another example: Povl Riis. The Q citation is for "Lois Reynolds; Tilli Tansey, eds. (2007), Medical Ethics Education in Britain, 1963–1993, Wellcome Witnesses to Contemporary Medicine, History of Modern Biomedicine Research Group, ISBN 978-0-85484-113-4, Wikidata Q29581753" but this is incorrect, the right citation is here. E.g. instead of "Wellcome Witnesses to Contemporary Medicine" it should be "Wellcome Witnesses to Twentieth Century Medicine". Normally, this is five seconds work. Now, I need to go to Wikidata, find the right item, change it: oops, I can't just change it, I need to find the right Wikidata item for "Wellcome Witnesses to Twentieth Century Medicine". It doesn't exist, so I need to create a new item. Or no, I should not do this, as apparently this is the older name for this Qitem[1]. But how can I make it display the older name? As far as I can see, I can't. So I have no idea at all how to solve this, bar going back to a standard citation instead of this Q thing.

Probably these things can be fixed by Wikidata-gurus, but the aim was to make things easier to edit, more accessible, not that we need to rely on a few experts to change these things. Fram (talk) 07:54, 12 September 2017 (UTC)

Clearly the main thing Wikidata needs at the. moment is more documentation and improved user interfaces. Is anybody working on these or at least trying to recruit volunteers for these tasks? —Kusma (t·c) 08:42, 12 September 2017 (UTC)

How would we know, and why would we care? You can ask at Wikidata, or you can ask the 5 or so editors here who have done 99% of the intrusion of Wikidata into enwiki. Fram (talk) 08:50, 12 September 2017 (UTC)

The correct place to raise this discussion is on the talk page of the template, not here (particularly when the main editor of that template has said "I'm done here" just above this). The way it works is quite straightforward: the line being used in the infobox has a reference on Wikidata of the type stated in (P248), which points towards the Wikidata entry for the entity being referenced (e.g. a book). Cite Q then fetches the reference information from the entry for the reference on Wikidata. If you want to edit which reference to use, then you use the Wikidata links in the infobox. If you want to edit the reference, you use the link in the reference - in the example at the top of this discussion, the link is at the end, as "Wikidata Q29581755". Thanks. Mike Peel (talk) 12:01, 12 September 2017 (UTC)

In this particular case it does not work however, and I raised the question at the Wikidata Project Chat.--Ymblanter (talk) 12:06, 12 September 2017 (UTC)

The correct place to raise this discussion is at a village pump, not at some hardly known template. Asking the template editors whether they want the template or see a use for it is hardly helpful, getting outside opinions is what is needed. As for this being "quite straightforward"; as shown above, not really. "If you want to edit which reference to use, then you use the Wikidata links in the infobox." If the reference doesn't exist on Wikidata, it is much easier to simply add it here than on Wikidata. I gave an example of an actual use of this template, where I want to change the name of the journal. However, the journal already has an item on Wikidata, but this is for both the old and new name of it. I shouldn't create another item (assuming I know how) for the same subject, but I can't use the alternative name either. There probably is some solution, but if this had been a regular cite template, it would have been straightforward and easy, instead of this. I have now replaced the source at my example with something useable. Fram (talk) 12:26, 12 September 2017 (UTC)

• Comment Meta data for refs need to go in articles. We could have a bot replace the "cite Q" with a fleshed out ref. A WD link can be added to cite journal if people want. Doc James (talk · contribs · email) 01:13, 15 September 2017 (UTC)
  • Why? That's certainly not the way I'm used to doing it in academic publishing (where the metadata goes in a separate bibliographic database from the main article text). Putting it in the article means that editors need to find another copy of it and copy it (propagating errors) or format it themselves (usually badly). Keeping it in Wikidata means that there is a single unified copy that can be cleaned up when it needs cleaning up. —David Eppstein (talk) 01:16, 15 September 2017 (UTC)
    • Making a major change like this requires consensus. I am not seeing that consensus. I am not happy to go hunting in another database for details to figure out what a ref is for. Also it is too prone to vandalism. This is why we did away with Template:Cite pmid. Here by the way is the discussion[2] Doc James (talk · contribs · email) 01:19, 15 September 2017 (UTC)
• The problem is ease of actual editing. When using this template, what we see in edit mode is a string of meaningless numbers like "Q123456", but what we expect to see (so we can edit) is the actual text... the name of the author, the book title, etc.

Perhaps the disconnect here is that those who work at wikidata think of citations as bits of data... which can be stored anywhere and retrieved when needed. Meanwhile those who work at Wikipedia think of citations as text... which needs to be located on the page, and visible in edit mode in order to edit it. Blueboar (talk) 00:18, 16 September 2017 (UTC)

But as I argued repeatedly elsewhere, we have the same problem for complex templates in Wikipedia, which sometimes make it more difficult to correct the data even than if the data is stored on Wikidata which is one click away from every page. This was like that already when I started, I have not seen any complaints, and nobody nominates these templates for deletion because of these usability issues.--Ymblanter (talk) 08:23, 16 September 2017 (UTC)

• have nominated for deletion: Wikipedia:Templates_for_discussion/Log/2017_September_15#Template:Cite_Q Jytdog (talk) 08:26, 15 September 2017 (UTC)

Mainspace use: subst?

Seems very well possible to use the template in mainspace with a subst:, thus:

{{subst:cite Q|Q29581755}}

Resulting in (without the TfD tag):

Lois Reynolds; Tilli Tansey, eds. (2008). Superbugs and Superdrugs: A History of MRSA. Wellcome Witnesses to Contemporary Medicine. History of Modern Biomedicine Research Group. ISBN 978-0-85484-114-1. OL 23194335M. Wikidata Q29581755., Wikidata Q29581755

The generated code is a bit unsightly (could be smartened up I suppose):

Collapsed 91 line wikitext expansion of {{subst:cite Q|Q29581755}}

{{#invoke:citeq|cite_q |qid=Q29581755 <!-- Simple parameters from Wikidata --> | publisher = {{#invoke:WikidataIB |getValue |qid=Q29581755 |P123 |fetchwikidata=ALL |onlysourced=no |noicon=true}} | oclc = {{#invoke:wd|property|Q29581755|P243}} | place = {{ #property:P291 |from=Q29581755 }} | doi = {{ #property:P356 |from=Q29581755 }} | issue = {{ #property:P433 |from=Q29581755 }} | volume = {{ #property:P478 |from=Q29581755 }} | pmid = {{ #property:P698 |from=Q29581755 }} | ol = {{#invoke:String|replace|{{#invoke:wd|property|Q29581755|P648}}|^OL(.+)$|%1|plain=false}} | arxiv = {{ #property:P818 |from=Q29581755 }} | bibcode = {{ #property:P819 |from=Q29581755 }} | jstor = {{ #property:P888 |from=Q29581755 }} | mr = {{ #property:P889 |from=Q29581755 }} | ssrn = {{ #property:P893 |from=Q29581755 }} | pmc = {{ #property:P932 |from=Q29581755 }} | lccn = {{ #property:P1144 |from=Q29581755 }} | hdl = {{ #property:P1184 |from=Q29581755 }} | ismn = {{ #property:P1208 |from=Q29581755 }} | journal = {{ #property:P1433 |from=Q29581755 }} | citeseerx = {{ #property:P3784 |from=Q29581755 }} | osti = {{ #property:P3894 |from=Q29581755 }} | biorxiv = {{ #property:P3951 |from=Q29581755 }} <!-- Special handling of Wikidata parameters --> | isbn = {{#if: {{ #property:P212 |from=Q29581755 }} <!-- if ISBN 13 --> | {{ #property:P212 |from=Q29581755 }} <!-- use ISBN 13 --> | {{ #property:P957 |from=Q29581755 }} <!-- else use ISBN 10 --> }} | others = {{#if: {{ #property:P110 |from=Q29581755 }} <!-- is there an illustrator? --> |Illustrator: {{ #property:P110 |from=Q29581755 }} <!-- use illustrator --> | <!-- else do nothing --> }} <!-- Optionally-local parameters --> | chapter = {{#if: <!-- is chapter specified? --> | {{{chapter}}} <!-- use locally-specified value --> | {{ #property:P792 |from=Q29581755 }} <!-- else fetch from Wikidata --> }} | date = {{#if: <!-- is date specified? --> | {{{date}}} <!-- use date --> | {{ #property:P577 |from=Q29581755 }} <!-- else use date from Wikidata --> }} | series = {{#if: <!-- is series specified? --> | {{{series}}} <!-- use series --> | {{ #property:P179 |from=Q29581755 }} <!-- else use series from Wikidata --> }} | title = {{#if: {{ #property:P1476 |from=Q29581755 }} <!-- is there a title --> | {{ #property:P1476 |from=Q29581755 }} <!-- use title --> | {{#if:{{ #property:P856 |from=Q29581755 }} |{{#invoke:WikidataIB |getLabel | Q29581755 }} <!-- don't mix |url= and a wikilinked |title= --> |{{#invoke:WikidataIB |getLink | Q29581755 }} <!-- else use label from Wikidata --> }} }} | url = {{#if: <!-- is url specified? --> | {{{url}}} <!-- use url --> | {{ #property:P856 |from=Q29581755 }} <!-- else use url from Wikidata --> }} <!-- Non- Wikidata parameters --> | accessdate = | author-mask = | df = <!-- this should eventually be superseded by a user preference setting --> | display-authors = <!-- redundant until multiple-author issue is resolved --> | embargo = <!-- for PMC --> | id = | mode = <!-- this should eventually be superseded by a user preference setting --> | page = | pages = | quote = }}{{#ifeq:{{{mode}}}|cs1||,}} [[Wikidata]] [[:d:Q29581755|Q29581755]]

--Francis Schonken (talk) 12:37, 19 September 2017 (UTC)

;-) Have you actually looked at the resulting code, or only at the screen output? I don't think many people will agree that the result you produce is any better, it seems to combine the worst of both worlds. Fram (talk) 12:42, 19 September 2017 (UTC)

Previous had an edit conflict, you obviously looked at the resulting code since :-) Still, the remainder of my reply stands. Fram (talk) 12:44, 19 September 2017 (UTC)

"a bit" unsightly was of course an understatement... Newbie user-friendlyness plummeting to a new unknown depth... Yeah, its current code output should not be invoked in mainspace by a subst:... --Francis Schonken (talk) 12:53, 19 September 2017 (UTC)

Holy crap. A proposal to BAN that would pass with SNOW consensus in a few hours. We would need a method that actually resolves all of the parser function calls like #if and #invoke and #property. Alsee (talk) 22:34, 21 September 2017 (UTC)

Concerns raised around Wikidata descriptions on English Wikipedia

Hey Folks,

There continues to be a lot of productive and helpful conversation going on. Thank you to everyone who has chimed in so far. I share Carwil’s sadness in feeling like there are bigger issues here that I hope we can solve them (last paragraph of this diff), but for this particular incident, I just wanted to summarize again what I’ve been hearing. This has been incredibly helpful to me and I am sorry that for some of you it feels like you’re yelling into cotton-filled ears. Your patience is appreciated.

Below are the main points I hear being raised as a response to my statement detailing what I thought the highest priority issues and potential solutions were.

Some caveats:

• Some folks called out that they thought Wikidata descriptions were okay, or disagreed with concerns below, but I didn’t include those points and instead focused only on the concerns.
• I also excluded the back and forths about relative quality of Wikidata. It seems that all of us can acknowledge it is a concern for some and that there is disagreement about how this does or does not impact Wikipedia or whether the proposed measures would sufficiently address concerns.
• Below are the concerns I heard, I am not endorsing those concerns, but stating their essence as heard.

Heard Concerns:

Perception that the Wikimedia Foundation overstepping its authority by interfering with content

• Placing perceived content at the top of each page, without editor consent is problematic since editors cannot easily moderate it.
• Perception that this is a violation of the Foundation’s terms of service and it’s role in the movement.
• This is about process and principle. The Wikimedia Foundation did not gain consensus and it likely would not have for the same reason.

Problem with making it easier to edit and track

• Wikidata editors can still revert a correction and it is unclear what an EnWik editor can do without getting involved in another project’s administrivia
• Some (or all?) enwiki editors have no interest in working on Wikidata’s policy
• Some (or all?) enwiki editors have no interest in editing Wikidata or making it better
• The Wikimedia Foundation missed that protected pages should have protected descriptions
• Based on above, the Wikimedia foundation is unlikely to be successful if they just do the suggested upgrades and hold an RFC for it

The Commons analogy

• The belief that Wikidata is not like Commons, for the following important reasons:
  • Smaller community
  • Less restrictive policies
  • Wasn’t made for Wikipedia ingestion
• There is a difference between choosing to put an image link in your article and having the Wikidata description placed at the top without your agreement
• Others seem to suggest that the issues are not fundamental or unfixable.

Concern about quality on Wikidata

• Perception that the impact of Wikidata vandalism unacceptably lowers Wikipedia's quality

A potential solution proposed

• Auto-create local Wikidata descriptions from the first sentences of articles

Special thanks to those of you who were clearly trying to bridge across the different perspectives, while being emotionally generous, kind and polite. This really is hard for the folks who worked on this at the foundation and for the English Wikipedia contributors who feel strongly, so your approach goes a long way.

We’re going to have to figure out the next steps together and we will be reaching out soon to move this along. In the meantime, I just wanted to let you know that we are listening and taking this very seriously at the Wikimedia Foundation.

Thanks, Jkatz (WMF) (talk) 22:13, 18 September 2017 (UTC)

I would only like to add: Auto-create local Wikidata descriptions from the first sentences of articles - alternative: obtain descriptions from a specialized template like {{Description}} in the article header or body. Thanks, —PaleoNeonate – 22:21, 18 September 2017 (UTC)

That is a decent summary. Thanks for hearing. Jytdog (talk) 22:34, 18 September 2017 (UTC)

I think it's worth noting that the views being expressed on this page are somewhat biased towards an anti-Wikidata viewpoint. The discussions on this page aren't really conducive to having a neutral conversation as pro-Wikidata voices (to balance the POV) aren't particularly welcome, and I'm not sure that people in the middle are represented here. I think it's good to pay attention to the concerns, and resolve them as much as possible, but please don't think that everyone on enwp thinks this way. Thanks. Mike Peel (talk) 23:57, 18 September 2017 (UTC)

I have to concur that this is a hangout of the anti wikidata crowd which does not reflect the complete enWiki editorship. Most of the editors that would like to use data from Wikidata will only find out about this when their access to Wikidata gets blocked through deletions or decisions made by a small group. Agathoclea (talk) 07:59, 19 September 2017 (UTC)

Well, the last RfC we had on the use of descriptions was not here but on the village pump. It seems to me that each time some use of Wikidata (the descriptions, Cite Q, ...) gets more exposure, outside the small circle of ardent Wikidata includers here, there is suddenly much more opposition. While this page probably doesn't reflect the enwiki community, just like template pages don't, it seems unwise to dismiss the criticism and discussion here as simply "the anti wikidata crowd" when actual, larger discussions show that the support for Wikidata on enwiki is seriously lacking outside this page and circle as well. Fram (talk) 08:15, 19 September 2017 (UTC)

Mike Peel I find it ironic you claim an anti-Wikidata bias here. You have over 16k wikidata edits, and I think you'd be hard pressed to find generic discussions on Wikipedia with a comparably high level of involvement by people with hundreds, thousands, or tens of thousands of wikidata edits. This page has attracted very motivated parties on both sides of the topic, as commonly happens with localized discussions. That's why RFC's actively call in uninvolved participation, and why we run RFC's like this at Village Pump. The previous RFC was clearly heading for removal of current wikidata descriptions, and I think it was leaning against the kind of upgrades proposed here. If the most involved parties this page fail to sort out an agreeable resolution, a new Village Pump RFC can sort out the true "unbiased" position. Alsee (talk) 01:38, 22 September 2017 (UTC)

@Jkatz (WMF): please note that this is a general page about Wikidata on enwiki: I deliberately started separate sections about e.g. vandalism or sourcing on Wikidata because they are not the same issue as the use of Wikidata descriptions in views, searches, ... Problems with other aspects of Wikidata on enwiki (see e.g. Wikipedia:Templates for discussion/Log/2017 September 15#Template:Cite Q, which gives some good insights in how people here feel about Wikidata on enwiki in general) should not be conflated with the singular issue of the descriptions imposed by the WMF on enwiki, and I would urge the WMF to first solve this one issue before moving on to the more general concerns and discussions. Fram (talk) 07:00, 19 September 2017 (UTC)

@Jkatz (WMF): I think you missed one "concern about quality in Wikidata". It's that Wikidata contains millions of statements that are not reliably sourced, according to en:Wikipedia standards. These statements are either unsourced, sourced from Wikipedias (which are not considered reliable sources in Wikipedia), or cite sources deprecated here.

It has to be acknowledged that Wikidata contains statement types that don't need sourcing or can't be meaningfully sourced (say, a Commons category). That skews statistics like this one, against Wikidata.

And it also has to be acknowledged that Wikipedia (incl. en:WP) contains millions of statements that are unsourced, many of them in legacy articles written years ago when referencing standards were less rigorous. I reckon well over half of all statements present in Wikipedia articles (article bodies rather than leads) today lack a footnote citing a reference.

But Wikipedians by and large do make an effort to ensure that any new content added to pages – especially the more visible pages – is sourced. They spend hours arguing over whether a particular source is reliable or not. They've thought about it over the years, educated themselves about who publishes what and with what level of professional oversight, consult and refer to precedents in places like WP:RSN, and they are proud of what they've learned since they started contributing, and the culture they've built.

And now you're coming with a project where millions of database statements that would be relevant to Wikipedia are either completely unsourced or sourced as "imported from Italian Wikipedia", "imported from Lithuanian Wikipedia", "imported from Cebuano Wikipedia", etc., without so much as a link to the article version that originally contained the info, let alone the source (if any) cited in that Wikipedia article at the time the import was made. And Wikipedians are supposed to be happy about content from that database being added to Wikipedia in various ways they can't control, when it hasn't been built to their sourcing standards. Can you understand that that goes against the grain and feels like a step backwards, diametrically opposed to the learning curve each of them has undergone to assimilate the culture that has been built here? I think there would be far less resistance to Wikidata if the sourcing standards matched. --Andreas JN466 14:33, 19 September 2017 (UTC)

I am one of those supporting the principle of using Wikidata in Wikipedia, but I must agree with Andreas here: the referencing of claims to Wikipedia articles in Wikidata is awful. There is no reliable source, and there is not even a date given when that Wikipedia might have contained the information. —Kusma (t·c) 15:05, 19 September 2017 (UTC)

Actually, these data should not be there at all. Indeed it was easy to take statements from Wikipedia(s), however, effort must have been made to also use the sources Wikipedia uses (and if these do not exist, not to import the statement). Most of them can still be easily fixed: if the item on Chad contains a statement that this is a country and the continent of this country is Africa, both being sourced to Wikipedia or unsourced, it is relatively easy to find a reliable source for both these statements and correct them by a bot.--Ymblanter (talk) 15:29, 19 September 2017 (UTC)