Talk:Reliability project

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search


Discussion of better tools for citations started at w:en:Template_talk:Ref_supports2, but it became clear that was the wrong place, so I've moved it here. Please join in. LeadSongDog (talk) 03:10, 20 March 2016 (UTC)

How adding citations/supported text should work in the editors[edit]

Thanks, LeadSongDog. I won't contribute here because looking at code gives me a headache. I'm happy to leave the technical design to WMF. That's what we pay them for. What I'd like us to do is work out a clear vision of how we want this thing to look and work - from the editor's perspective.

My preferred solution is: the editor pastes a doi, url, ISBN, PMID or whatever into a form (via the visual editor), adds page numbers, copy/pastes the article text being supported into the form, and clicks save. That's about as much typing as I think we should be expected to do. What do you think? --Anthonyhcole (talk · contribs · email) 01:35, 18 March 2016 (UTC)

Not everyone uses the visual editor, some of us prefer the wikitext. Either way, keeping the wikitext as simple as possible pays off. What bothers me about the existing template is too much extraneous detail, too hard to use. The ref tags serve no real purpose, nor the repetition of text. Hence my suggestions above. LeadSongDog come howl! 02:57, 18 March 2016 (UTC)
Definitely. If the wikicode can be minimised like that, all the better. --Anthonyhcole (talk · contribs · email) 08:11, 18 March 2016 (UTC)
First, I have split this into a separate topic, as it appears broader than just the specific syntax used for this template. I also wonder if there is a more appropriate place to have these broader discussions (this and the one I added below regarding how the support should be indicated to the reader) than the talk page of a fairly obscure template. We can move it elsewhere, or talk here for a while and then move it elsewhere for broader participation.
We should consider how it is desired to have editors indicate the connections between the reference and supported text.
Visual Editor:
I, also, don't use the visual editor. However, I have ideas about how indicating what text is supported could be input and displayed. Ideally, the user should be able to do something as simple as highlighting some text with the mouse, right-click and select "add supporting reference" (or click a button, or press a hotkey) then add any of the identifiers which Anthonyhcole mentioned into a popup and have everything filled out for them (there really should be a display of the citation which the user should check). Alternately they could right-click on a reference and select "associate supported text", or some such, then highlight text (or highlight, then right click on the reference and select "highlighted text is supported by this reference".
I would also want there to be a warning popup when someone is changing text that is specified as supported to force them to click on something indicating that they have verified that the new text is still supported by the reference. Perhaps, if it is an online reference, it should offer to open a tab with the reference for the user to look at.
The new syntax is better, but still cumbersome. With a version of indicating supported text that includes the use of JavaScript, or changes to the MediaWiki software, it would be possible to have parameters that indicate that the previous X sentences, or paragraphs, are supported by the reference. However, there is a reason to have the text that is supported be explicitly stated (duplicated). That reason is that text often gets changed by editors without checking that the reference supports the change. This can migrate, through years, to the point that what is in a paragraph with a reference has no relationship to what the reference actually says. In my opinion, there should be at least some type of necessary additional step that someone has to take in order to change text that is explicitly supported by a reference in order for the text to still show as supported.
An alternate method of indicating connections is to have the supported text enclosed in a template within the paragraph. This also becomes cumbersome when there are multiple references with overlapping supported text. An example of this method is used for the the third style of indicating supported text that was implemented, as mentioned below in #How, ideally, should the supported text be indicated to readers on Wikipedia pages. (Example page) implemented with {{Supported by ref}} and {{Ref supports3}}
— Makyen (talk) 08:53, 18 March 2016 (UTC)
Makyen, I like

"Ideally, the user should be able to do something as simple as highlighting some text with the mouse, right-click and select "add supporting reference" (or click a button, or press a hotkey) then add any of the identifiers which Anthonyhcole mentioned into a popup and have everything filled out for them (there really should be a display of the citation which the user should check)."

A lot. --Anthonyhcole (talk · contribs · email) 08:12, 19 March 2016 (UTC)
I've no objection to moving the discussion, it's just a bit early to figure out where to take it. There have been many such discussions over the years in many places. I'd like to avoid getting caught up in implementation details ahead of getting somewhere with the basic concepts.


Perhaps we can start by getting clear terms. It likely is sensible to adopt nomenclature consistent with Wikidata in order to facilitate translation efforts. I'll suggest this terminology:

An statement without citation "may be challenged" by the addition of the {{citation needed}} tag. A statement that seems to be possibly false-valued "may be challenged" by the addition of the {{verification needed}} tag. To verify that a citation is correct, it is only necessary to locate the corresponding bibliographic record in a trusted catalog. Inability to locate such a record should trigger addition of a {{citation error}} instance. To verify that the statement is true, it is necessary for a reviewer to locate and obtain another instance of the reference work, find the relevant part of that work, and compare it to the statement being tested. Having done so, if the citation does not support the statement it should trigger a {{fails verification}} instance. If it does support it, there should someday be a mechanism for the reviewing editor to so indicate against a specific revision of the statement, or equivalently a revision of the encyclopedic article. Occasionally, the difficulty of accessing a copy of the reference work means that it must be left up to a wikignome who cannot understand its meaning, whether for reasons of language, or technical difficulty. Such cases often crop up at wp:RX. These may lead to a substantial section of the work being copied and sent to another editor who can understand it and perform the verification. It is sometimes helpful if these verifiers either quote the relevant bit or store a copy in case further examination is necessary. The scarcer the source the more helpful this approach can be. The verifying editor may use the {{quotation needed}} markup to request this. LeadSongDog come howl! 19:37, 18 March 2016 (UTC)
I agree with all of that, and agree with Makyen that this conversation deserves preserving on somewhere more broad-stroke than a template discussion page. I've just created . What do you say we move this there? --Anthonyhcole (talk · contribs · email) 08:01, 19 March 2016 (UTC) Yes check.svg Done

What is under discussion?[edit]

Bibliographic record reuse[edit]

Within an encyclopedic article, a given reference work may be cited many times in support of different statements: we need to distinguish the various cited parts (whether by "pages" or "at" or "paragraph") from the reference work as a whole. As articles are interwikied or translated, a given reference work may be cited at many places through many wikis. Replicating the chore of citation in all these places is wasteful of valuable editor effort. All of these places reusing the content could benefit from one fully-developed bibliographic record to draw upon. We may characterize such a record simply by the clarity with which it identifies the specific reference work. It should contain one or more robustly supported bibliographic identifiers, such as an ISBN, an OCLC number, a DOI, or a PMID, such that any competent reference librarian could tell what the cited work is. It should identify the language of the reference work. It could eventually be translated into the language of each wiki where it is used, although this is a secondary goal. Backing up transient URLs with persistent URIs to archived copies is a long-recognized need, as is the need to replace naked URLs with more complete references in order to preclude w:en:Linkrot.

Local attempts to enable reuse were made at w:en:template:cite pmid, but were not sufficiently thought through. By putting them in template-space on one wiki, they were rendered difficult to reuse elsewhere. Instead, they should reside in a structured bibliographic database that can facilitate internationalization, localization, and repurposing. Tools will be needed for each wiki to draw upon that database and render locally-accepted citation formats. Similarly, tools will be needed to turn locally-created citations into nonredundant database records. Deduplication of records can be based on multiple criteria, and should be largely automatic.

Where used[edit]

For each bibliographic record, it should be possible to list all the articles on all the wikis that cite the corresponding work. Further, it should be possible to flag the existence of retractions and updates, so that editors of articles citing an edition can be alerted to the existence of more current information. This might simply be an "updated-in" link.

Prior art[edit]

A substantial attempt at much of the above was made at the w:en:Open Library, but it has long been badly underresourced, so that the present state of the software there is quite disfunctional. One lesson to learn from it is that authority files (for authors and publishers) must not be free-text, as that leads to near-duplicate records which require manual correction. Instead, a controlled repository such as the VIAF should be used. This should be simple, now that authority control data is resident in Wikidata, and corrective processes are in place at w:en:wikipedia:VIAF/errors with the help of @AndyMabbett:.

Most library catalogue records are represented in structured forms, such as MARC-21. Originated in the punch-card era, these are information dense, but difficult for untrained humans to read. Newer formats exist, including both open and proprietary ones. A great deal of expert effort has gone into developing tools for catalogue records management that can surely be leveraged.

The world's largest bibliographic catalogue is a virtual union catalogue, w:en:WorldCat, coordinating, merging, and sharing records among a large proportion of the world's libraries. It is neither complete, nor entirely open, but it is by far the best freely-usable tool available for finding information about unfamiliar works. Full participation is reserved for paying member institutions, but there has in past been a Wikipedian in Residence there, @Maximilianklein:.

I now see that a substantial effort seems to be underway at WikiCite 2016, so perhaps this is all redundant.@DarTar:

Still, let's talk. LeadSongDog (talk) 21:07, 22 March 2016 (UTC)