Wikipedia talk:WikiProject Short descriptions/Archive 3

This is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

Following a request made on WP:US/R, a new script has been written (User:SD0001/cat-all-shortdescs.js) using which you can see at a glance the short descriptions of all pages in a category (well, not all, but of the 200 which are in view at a time). SD0001 (talk) 17:37, 21 August 2020 (UTC)

User:GhostInTheMachine/SDlinkBuilder may also help. It helps you make a short description check list for all of the links in any page. — GhostInTheMachine ^{talk to me} 18:47, 21 August 2020 (UTC)

SD0001 Could this be adapted to check the short descriptions for articles in a Wikiproject where the talk pages are tagged, like Category:WikiProject SCUBA articles? · · · Peter Southwood ^(talk): 19:54, 26 August 2020 (UTC)

A very useful tool, that would be even more useful if extended as described for project maintenance. · · · Peter Southwood ^(talk): 05:05, 27 August 2020 (UTC)

@Pbsouthwood:  Done Amazingly, this also makes the script faster, even though it now has to do more API calls! SD0001 (talk) 07:22, 27 August 2020 (UTC)

@SD0001:. Great stuff, that was quick;-) Cheers · · · Peter Southwood ^(talk): 07:40, 27 August 2020 (UTC)

The colour coding is a huge help. · · · Peter Southwood ^(talk): 08:24, 27 August 2020 (UTC)

@SD0001: Where exactly does the button appear? I don't see it, either in Firefox or Chrome on a Mac. I am using RedWarn, which has buttons at the top on the right. Could those be interfering? MichaelMaggs (talk) 09:58, 27 August 2020 (UTC)

@MichaelMaggs: Just under the Pages in category "xxx" header, towards right. SD0001 (talk) 10:01, 27 August 2020 (UTC)

Nope, nothing. Have reinstalled the script, but still only see the standard wording "The following xxx pages are in this category, out of xxx total. This list may not reflect recent changes (learn more)." MichaelMaggs (talk) 10:17, 27 August 2020 (UTC)

That's odd. This is how it should be like. I've checked with redwarn, it shouldn't be interfering. And after u click it, things should become like this. SD0001 (talk) 11:55, 27 August 2020 (UTC)

Very nice and clear in your screenshot, but not on mine unfortunately. I wondered if it might be to do with the recent Vector skin updates, so I went back into Preferences | Appearance and switched on the Use Legacy Vector option. But that has no effect. Oh well. MichaelMaggs (talk) 14:06, 27 August 2020 (UTC)

MichaelMaggs just seeing this now. Check if there are any errors in the console (WP:JSERROR #6). SD0001 (talk) 04:23, 29 August 2020 (UTC)

SD0001 Not really sure what I'm looking at, but the top line (in red) says "Uncaught ReferenceError: button is not defined, which can be expanded to give 10 or so lines of detail the first of which reads https://en.wikipedia.org/wiki/Category:World_Heritage_Sites_in_England line 11 _ injectedScript:1. Then two GET errors relating to arrow-expanded.svg and arrow-collapsed-ltr.svg. Maybe here isn't the best place to discuss in detail, but I’m happy to help if you want more information or if you need some tests to be carried out. Let me know where the best place to discuss would be. MichaelMaggs (talk) 17:41, 29 August 2020 (UTC)

I have made a list at Wikipedia:WikiProject Short descriptions#Tools, gadgets and scripts. You might want to polish up the descriptions a bit. · · · Peter Southwood ^(talk): 05:25, 27 August 2020 (UTC)

I have no personal experience doing short description imports, so I'm seeking opinions on this kind of high-speed importing done by Feloniii. Is this an indication of somebody just trying to rack up their edit count as quickly as possible, or is the process really mechanical enough that 5 imports in 25 seconds could be a reasonable use of the tool, exercising an appropriate level of diligence? -- RoySmith (talk) 17:39, 24 August 2020 (UTC)

Now blocked, it seems. But in principle such quick editing could be done in short bursts if the editor already has many tabs open and goes through them one by one, accepting the Wikidata text in each case. That's just one click to move to the next tab, and one click to import. Usually, though, it's not possible to do more than a few before the text needs to be edited. MichaelMaggs (talk)

At the risk of opening my edits to scrutiny, I have had no trouble doing 50 short description imports in 14 minutes, with runs of 8–10 per minute in cases where the need to modify the SD was minimal or absent. I was working from a list, and I open 50+ tabs, wait for them to load, and then click through each tab, using the SD Helper to import or edit (or skip, closing the tab) each SD. – Jonesey95 (talk) 20:56, 24 August 2020 (UTC)

That looks totally plausible for reasonable quality Wikidata descriptions and reasonably clear lead sections. · · · Peter Southwood ^(talk): 15:55, 31 August 2020 (UTC)

• OK, I'm going to assume this is nothing to be concerned about. Thanks for the input. -- RoySmith (talk) 02:40, 25 August 2020 (UTC)

When importing and browsing through short descriptions, I've come across some instances where, say, a town will have "Town in Los Angeles County, California, United States", whereas another might have "Town in Los Angeles County, California, U.S.", or just "Town in Los Angeles County, California". Is there a consensus on which of these is correct? I've seen a similar pattern with UK places. Sdkb (talk) 17:15, 5 July 2019 (UTC)

Since that specific template is probably {{Infobox settlement}} it already has the code to create a consistent short description (whichever that is). If you different styles, that probably means someone added a manual one, which might not be helping at all. As for your question - "United States" and no "U.S.", but again, that should be covered by the infobox. --Gonnym (talk) 17:55, 5 July 2019 (UTC)

Coming back to this a year later, to clarify the scope of my question, I'm wondering not just about place articles, but also e.g. "Building in Springfield, Oregon, United States", "Museum in Springfield, Oregon, United States of America", or "College in Oregon, United States". Not all of those instances are defined by an infobox, and not all of the infoboxes that do define are consistent. {{u|Sdkb}} ^talk 22:19, 12 September 2020 (UTC)

WP's Manual of Style is pretty clear on this, MOS:US. My reading is both "US" and "United States" are allowed styles, at each editor's discretion. I prefer "US" in a short description, but it should not be the only style allowed for short descriptions. What is not style (quoting from MOS:US): "Do not use the spaced U. S. or the archaic U.S. of A., except when quoting; and do not use U.S.A. or USA except in a quotation, as part of a proper name (Team USA), or in certain technical or formal uses (e.g., the ISO 3166-1 alpha-3 codes and FIFA country codes)".

A consensus on this web page doesn't override the MoS and will have little effect on the rest of WP editors. See the relevant pages in the Wikipedia namespace for how and where to achieve such consensus & change the MoS. — Lentower (talk) 14:35, 13 September 2020 (UTC)

Allowing a fairly wide degree of variation in normal article prose is fine, but since short descriptions are more tightly controlled, I think it's not ideal if the Getty uses "Art museum in Los Angeles, California, United States" but LACMA uses "Art museum in Los Angeles, California, U.S." Short descriptions are a more restricted environment in which we can and should aim for a higher degree of standardization. So long as it's consistent with one of the options in the MoS, I don't think we have a problem. {{u|Sdkb}} ^talk 15:55, 13 September 2020 (UTC)

I support more standardization here also. I manually added one today "School district in Illinois, U.S." which is 33 characters. Within about an hour, someone changed it to "School district in Illinois, United States" which bumped it a bit over 40. Both are MOS compliant and being a bit over 40 is not terrible. But U.S. is well, short, which seems to me to be more consistent with the intent of a SD. I wouldn't change it back without a guideline that says appropriate abbreviations are preferred. MB 03:26, 14 September 2020 (UTC)

I have been told in Wikidata that my descriptions there don't follow their conventions (start lowercase). I was unaware that I was contributing there, but it seems that sometimes when I use the Wikipedia:Shortdesc helper to edit an en:wikipedia article, somehow my edit ends in Wikidata as well. It does not happen in other cases. What am I doing in the short desc helper that writes in Wikidata? Is the Edit and import? --Error (talk) 22:40, 12 September 2020 (UTC)

Something appears to be incorrect in our documentation, which says Adding a short description on Wikipedia does not do anything on Wikidata. I see that I also have been inadvertently adding short descriptions on Wikidata.

If you click on the Edit link for a short description, then choose the gear wheel, you should see an option that says "Save changes to Wikidata". The default, in apparent contradiction of this page's documentation, says "Only when no Wikidata description exists (default)". The other option is "Never".

As for the capitalization, I have always assumed that a bot will be able to take care of the conflicting capitalization. The last time I checked, Wikipedia's guideline was to capitalize the first letter (except in rare cases like "iPod"), while Wikidata's guideline was to lower-case the first letter except for proper nouns. – Jonesey95 (talk) 23:50, 12 September 2020 (UTC)

Copying correctly from Wikipedia to Wikidata assumes that the bot knows a proper noun (other than an iPod) when it sees one. Certes (talk) 00:07, 13 September 2020 (UTC)

Sorry if I was unclear. When I referred to a bot, I meant a bot operating only on Wikipedia, or only on Wikidata, that would tidy up capitalization of descriptions. If not a bot, an intelligent editor with a script, operating from a report of likely miscapitalizations. – Jonesey95 (talk) 00:22, 13 September 2020 (UTC)

The right way to use the shortdesc helper to add a description is to put the first char as lowercase (unless it's a proper noun). Shortdesc helper will automatically uppercase the first char while saving it to wikipedia, but post the unaltered description to wikidata. Voila, now we're compliant with both projects' policies! Maybe we should document this hack somewhere? – SD0001 (talk) 03:00, 13 September 2020 (UTC)

I wasn't aware of that. And yes, it would be good to have that feature documented. That doesn't work, though, in the majority of cases where Wikidata already has a Short Description. In such cases, editing the text with the gadget does not update the Wikidata text unless the Export button is used after saving. Unfortunately since the saving step always converts the first letter to upper case, there's no way to comply with Wikidata's policy of starting with lower case. So far as I know there is still no way to correctly update an existing Wikidata SD without going over there and doing it separately. MichaelMaggs (talk) 10:27, 13 September 2020 (UTC)

Hmm yeah. That could be fixed by having a "Save and export" button in addition to the save button while editing an existing description. – SD0001 (talk) 10:47, 15 September 2020 (UTC)

@SD0001: That would be perfect! Or alternatively the suggestion that RexxS made here. MichaelMaggs (talk) 12:08, 15 September 2020 (UTC)

@MichaelMaggs: Looks like we may need to have a discussion to reach a consensus on these options. Galobtter doesn't seem to be very active at the moment, so I can try to implement it in the gadget once it's clear what option(s) should be added. – SD0001 (talk) 16:09, 17 September 2020 (UTC)

@SD0001: sure. I don't mind which option, as either would be a great improvement on what we have now. Where do you suggest the discussion should take place? RexxS may well be able to comment on the pros and cons. MichaelMaggs (talk) 16:46, 17 September 2020 (UTC)

Saving only uppercases on initial addition and not on subsequent edits (so if something really needs to be lowercase it can be set so). I guess the save and export could uppercase for Wikipedia and update Wikidata unmodified.

Export (lc) as suggested by RexxS seems a good option too. Implementing any of these options should be very easy and I should be able to do it if there's agreement on which option is best. Galobtter (pingó mió) 19:51, 17 September 2020 (UTC)

Please could we avoid the word "export". (You export data from a database, which could be seen to imply that the flow is from Wikidata.) If the SD is being copied to Wikidata, then please just say "Copy to Wikidata". — GhostInTheMachine ^{talk to me} 20:16, 17 September 2020 (UTC)

My understanding is that the Wikidata folks are ok with new descriptions being of the wrong case since that's better than no description, so that's the only case when shortdesc helper by default adds descriptions. The documentation definitely needs updating. Galobtter (pingó mió) 19:57, 17 September 2020 (UTC)

I have proposed that user:SD0001/shortdescs-in-category be made a gadget so that it can be found easily by more users. See Wikipedia:Village pump (technical)#New gadget_proposal: view short descriptions in categories. Please comment there if you find this script useful. Thanks, – SD0001 (talk) 05:07, 20 September 2020 (UTC)

Hi, I'm looking through a lot of short descriptions at the moment (importing to Wikidata where feasible and there isn't one there). Some odd ones are appearing, I'll highlight them in this thread if that's OK, hopefully someone here can help fix them.

• 2001–02 FK Partizan season has a short description of "Partizan 2001–02 '"`UNIQ--ref-00000000-QINU`"' football season"

That has been fixed. It was due to a value in the infobox, used in the SD, having a ref tag. I see this occasionally with school articles too. Just delete the ref or move it to the prose as appropriate. MB 14:15, 21 October 2020 (UTC)

@MB: Thanks - but this seems to be the case for most (all?) of the articles in Category:FK Partizan seasons. Thanks. Mike Peel (talk) 15:50, 22 October 2020 (UTC)

Mike Peel, Jonesy95 has started Wikipedia talk:Short description#Detecting problems in short descriptions. MB 17:40, 22 October 2020 (UTC)

• There are a number of cases where the short description is the same as the page title, e.g., 17th Congress of the Philippines, 1859 Boston mayoral election. This doesn't seem helpful/correct. (spotted as Wikidata doesn't allow descriptions and labels/article titles to be the same).

(... more to come...) Thanks. Mike Peel (talk) 08:00, 21 October 2020 (UTC)

It looks like there are about 200 instances of the former error in articles transcluding {{Infobox school}}. I have raised the issue at Template talk:Infobox school. – Jonesey95 (talk) 15:37, 21 October 2020 (UTC)

• Trang Ý has a short description '''Empress Lệ Thiên''' - surely wikicode shouldn't be in short descriptions?
• Matt Heaton has a random 🏉 in the short description

Thanks. Mike Peel (talk) 16:43, 21 October 2020 (UTC)

• The 7 Beer Itch seems to have a short description despite being a redirect? Thanks. Mike Peel (talk) 18:49, 22 October 2020 (UTC)
  • Gonnym might be able to explain the logic behind that one. – Jonesey95 (talk) 19:21, 22 October 2020 (UTC)
    • There is no reason redirects shouldn't have short description. This has been discussed here before and mentioned on the project page. --Gonnym (talk) 22:23, 22 October 2020 (UTC)
• Barbados Defence Force Sports Program has a short description of 'C'. Thanks. Mike Peel (talk) 10:55, 24 October 2020 (UTC)
• There are also some cases of single-word descriptions, e.g., Beer in Iceland has a description of 'Iceland'. Thanks. Mike Peel (talk) 13:44, 24 October 2020 (UTC)

From https://en.wikipedia.org/wiki/Wikipedia:WikiProject_Short_descriptions#Editing_using_Shortdesc_helper_gadget

Adding a short description on Wikipedia does not do anything on Wikidata. To edit on Wikidata, click on one of the links (not shown in this example) in the Wikidata item description or Q-number to open the item at Wikidata.

This is not true, actually. Turns out that it Does affect Wikidata! I noticed this when a user reverted many of my edits that were on Wikidata: https://www.wikidata.org/w/index.php?target=AltoStev&namespace=all&tagfilter=mw-reverted&start=&end=2020-10-16&limit=50&title=Special%3AContributions

Notice how all the edits show "Shortdesc helper".  AltoStev ^Talk 22:06, 17 November 2020 (UTC)

We have too many pages that try to cover the same topic. I have fixed the text in question by copying the correct explanation from Wikipedia:Shortdesc helper. – Jonesey95 (talk) 22:39, 17 November 2020 (UTC)

Nearly six 3.7 million Wikipedia articles still don't have any short description, and they should. Someone populated the Wikidata texts using a fairly simple bot, and there's no reason why the Wikipedia community here couldn't do the same, but with more sophistication and focusing on Wikipedia's specific needs. There's already an active group of editors working on this WikiProject. With a suitable commitment from a skilled bot writer, I'm confident there would be enough eyes to ensure a high level both of reliablity and of usefulness for our purpose. It takes a huge amount of time to edit and check each new SD manually if the article has to be opened each time, but if a bot writer could provide draft texts in a table format for manual review before updating, the manual effort would be far less. I, and I'm sure several others, would be only too happy to help with the technical specifications of such a bot, though I'm not able to do the actual coding. MichaelMaggs (talk) 13:44, 7 August 2020 (UTC)

I'm not sure where you got six million; I think it's more like four million (6 million articles, minus 2 million articles in Category:Articles with short description). In any event, I think PearBOT has been coded to do just this. It looks like that bot has processed some articles but has not done any work recently. It might be worth asking Trialpears whether that bot has done everything that it can do, or whether it has been paused for some other reason. – Jonesey95 (talk) 16:31, 7 August 2020 (UTC)

Sorry, yes, I misread the number. Corrected above. MichaelMaggs (talk) 16:39, 7 August 2020 (UTC)

(edit conflict) The bot can still do some more, but the project got kind of disrupted by some personal events as well as a pandemic. I can probably get it going again this weekend. The bot was based on the first sentence of each page with a lot of processing resulting in descriptions such as English Author or Portuguese Pianist for biographies. I think the same method could be used a lot more widely, but I got quite bored of it and find it very unlikely I will try to review several thousand generated descriptions for another similar bot. If anyone with the time and skills want to try I'm happy to give them some pointers and code though. --Trialpears (talk) 16:39, 7 August 2020 (UTC)

Many thanks. May well ask for pointers as we go on. MichaelMaggs (talk) 10:07, 8 August 2020 (UTC)

Miniproject to generate new short descriptions by category

I propose a concerted effort to work though the 3.7 million Wikipedia articles that still don't have any short description, using bot-assistance on a category-by-category basis to reduce the number to a bare minimum of difficult cases that can only be handled manually. Start with the remaining larger and most structured categories to sweep up as much easy stuff as possible, adding more complex criteria as we build up experience. All subject to manual checks of at least a significant sample of the cases.

Proposed procedure

1. Select a high-level category to update, the more structured the better
2. Consider only those articles that have no SD
3. Define trial criteria to pull out suitable articles, and define trial SD definitions for those articles
4. Tabulate all (or a sample) of the selected articles with their proposed SDs; and separately list all the articles failing the criteria
5. Review these manually, and repeat 3 and 4 to reduce errors to an acceptable level
6. Update the articles with the new SDs, and repeat from 1.

The proposal will need someone to write the code to generate the review tables, and to run an approved bot to update the articles themselves. Defining the trial criteria and SD definitions would need to be done separately for each top-level category, and more editors could provide input as that wouldn’t need such a high level of programming experience. I know enough Python myself, for example, to be able to run and edit a script, and could certainly help out with steps 3 to 5, especially if someone would be willing to provide assistance if I get stuck.

Proof of concept

As a proof of concept, we might perhaps start with Category:English footballers, which has around 11,000 articles with no SD.

Proposed trial criteria:

• has a single main infobox (eg checked via Module:Is infobox in lead) based on {{Infobox football biography}}
• the first 30 words of the lead include is|was a|an [+ 0 to 5 words] English [+ 0 to 5 words] football

(Restricting the selection by wording in the lead is necessary, as the category is sometimes used for non-English players who play for an English club. Need to pick up eg "Was a former English professional football player").

• article does not have a local short description (added for clarification by Jonesey95)

Proposed SD definition:

• English footballer [dates]

Where [dates] means "(birthyear–deathyear)", where known, extracted from birth_date = and death_date =. If only one is known, then (b. birthyear) or (d. deathyear)

Comments would be welcome. Would be possible to seek bot approval for the overall concept, without having to submit the trial criteria each and evey time? Mike Peel, is this something you would be willing to help with? MichaelMaggs (talk) 10:07, 8 August 2020 (UTC)

Not sure why you need a bot if your criteria is checking the infobox. If we take your example of {{Infobox football biography}}, why not have it handle it, and then save everyone from 176k pointless edits? --Gonnym (talk) 11:20, 8 August 2020 (UTC)

That would be ideal where an infobox is sufficient on its own to define a good, detailed SD. But in many cases using the text of the lead in addition to the category allows for more detailed and focused wording. It also allows better filtering of unexpected results, as here where a few editors have used Category:English footballers to mean "plays for an English club" regardless of nationality. Once the main script is written – which has to be done only once – it becomes useful in a whole range of scenarios, with different choices of trial criteria depending on context. MichaelMaggs (talk) 12:10, 8 August 2020 (UTC)

If the infobox is not sufficient to define a good, detailed SD, then the project should be to enhance the infobox so that it is sufficient. If a bot can suggest data to add to the infobox then that is the better way forward — the infobox is enhanced and a good SD is the end result. — GhostInTheMachine ^{talk to me} 16:23, 8 August 2020 (UTC)

Well, the primary purpose of this project is to add large numbers of new SDs quickly, based on some pretty flexible tests depending on the articles being processed. In many cases there may be no infobox ar all, and the SD would need to be constructed from a combination of the text of the lead and the category, alone. Even where there is an infobox, that box has its own purpose which isn't as a general rule to generate or replicate the SD. If the proposed text of a new SD comes at least partly from the lead plus the category, requiring that to be pushed back into an infobox would cause its own set of problems, including the need to discuss with a wide variety of infobox maintainers where that information should be held. While enhancing infoboxes could well be a focus of a project of this type, it's not something that is feasible to incorporate into this one. MichaelMaggs (talk) 18:17, 8 August 2020 (UTC)

• @MichaelMaggs: Seems straightforward enough. Here's the test code, and the results from running it through the first 100 in the category are at User:Mike Peel/shortdesc. It checks for "nfobox" in the text, but it doesn't currently check the first words in the lead. It also doesn't currently do any editing - it could either do that directly, or I could write a second script to download the list after manual checking and then make the edits, if you'd prefer. Thanks. Mike Peel (talk) 15:33, 8 August 2020 (UTC)

@Mike Peel: Many thanks Mike, that was really quick! Let's see what the general response to the idea is, now. I think the script will ultimately need a section of (drop-in replaceable) code to look at the text of the lead, not so much for this particular example but so that the script's operation can be tweaked easily for a variety of other situations, including where there is no infobox at all. All the tests for the trial criteria could then generate a separate exceptions output list, allowing the 'failed' articles to be checked manually - omitting entirely all articles that already have a short description as they are out of scope here. The separate main list would then just include the articles that pass the tests, plus their proposed SDs. Then, once checked, that list would be read back in by a second script to make the actual edits. MichaelMaggs (talk) 17:53, 8 August 2020 (UTC)

All sounds doable, I can code more of this up if the community wants this. Thanks. Mike Peel (talk) 17:57, 8 August 2020 (UTC)

Mike Peel, this looks like a great start! I have a couple of comments: 1. Can you please use a dash (–) between years, per MOS:DASH? 2. I don't think we need articles with local short descriptions to appear on the list at all. 3. Are you able to see the Wikidata short description, so that an editor could simply import it, as we can do with the Shortdesc Helper gadget?

Per the comments of Trialpears below, I would like to see a way to generate a list of 100 articles from a category name that I supply, view the list, and then have an "import" or "create" link to the left of each one (I suggest left, because otherwise the links would not be aligned vertically, unless you want to build a table) that would import the Wikidata short description or create a new one based on the bot's suggestion. As a bonus feature, an editor could provide a starting letter of the alphabet so that I could get a list of English footballers starting with "L" and beyond; this would allow me to skip tricky articles and still get a new list of 100 articles.

I don't think we need a bot to do a few thousand short descriptions at a time with manual oversight, although we might want to ask at BRFA if we would be contravening WP:MEATBOT. – Jonesey95 (talk) 19:01, 8 August 2020 (UTC)

The dash is easy. The rest would also be easy on Wikidata (just feed them into QuickStatements), but for enwp it's more tricky. I can try to figure out oauth, but even then, setting up an interface for others to run this would take time. The best I could easily offer in this case is tidy Python code that you could configure and run yourself. Thanks. Mike Peel (talk) 19:25, 8 August 2020 (UTC)

@Jonesey95: I'm pretty certain that pageprops contains the Wikidata description in the "Central description" field of [Page information] which the bot is already reading, so it should be immediately available unless I've missed something. --RexxS (talk) 19:27, 8 August 2020 (UTC)

As a quite general comment about bot made short descriptions: If there is a human in the loop approving them there are many hundreds of ways to generate them, but if you wish to make it fully automatic you will probably have to check many hundreds to several thousand descriptions before it is good enough. I think I've checked between 2000 and 2500 descriptions for PearBOT 5 and I've had several other people doing checks as well. This proposal is almost certainly fine with a human in the loop, but I'm unsure if you can get it good enough for fully automatic. --Trialpears (talk) 18:08, 8 August 2020 (UTC)

Yes, absolutely agreed. There will be too many to check each individual article manually, but the plan would be for me and hopefully other volunteers to manually check a good sample of each run, and I'd certainly expect that to involve several thousand checks. By repeatedly running with different focused rules for each category we can avoid the need to have a single set of overly-broad rules that has to fit everything. Do you have a feel for an error rate that might be achievable? I was hoping for, say, < 2%, but I'm not sure what's possible before digging into the results. MichaelMaggs (talk) 10:19, 9 August 2020 (UTC)

I tried to do this exact process about 2 years ago on articles about a species. The discussion can be found here. All it was in that case was simple regex to extract a chunk of the first sentence, do a little clean-up, and then apply a couple rules to see if the algorithm had likely botched it (too short, too long, etc). Ultimately we got it to the point where semi-automatic posting would've been fine, but fully automated posting would've been sketchy. Is it worth seeing if anything there is salvageable or if either of us did something unique and clever that can help the other one? Tazerdadog (talk) 22:43, 11 August 2020 (UTC)

@Tazerdadog: Thanks, that looked an impressive project and it was a shame there wasn't more interest at the time. My aim is to keep the error rate low by making use of the category name, and not attempting to do anything very clever based on the lead text. In many structured articles, in fact, a regex check against the lead would be needed only to catch categorisation errors (for example to avoid a false positive when a Scottish footballer playing for an English club has – perhaps incorrectly – been put into Category:English footballers. But as the plan is to construct a rather generic script that anyone could run, your approach could still be investigated, especially when we get to the more difficult unstructured articles. It would be great if we could collaborate. MichaelMaggs (talk) 12:55, 14 August 2020 (UTC)

Discussion of script specification

As Mike Peel has very kindly offered to write a script, we need to decide exactly what the requirements are. The script should be able to handle both bot and user-assist running. Bot operation will be needed if we are to make serious inroads into the remaining easy/structured cases in a reasonable time, and the ability to run as an assistive script as well will allow much faster manual editing than is possible with the existing gadget. Mike has said that he can't offer a complex user interface, which is fine. I would think that a simple Python file that anyone can run locally would be quite workable. The output/input mechanism needn't be pretty as long as it is accessible to anyone who knows how to run Python. Here's what I suggest; we should discuss before actually asking Mike to implement it

A Python script, based on the one already written, with the following enhancements:

1. Accept a user-defined category that, with all its sub-categories, defines the articles to be considered. In the initial trial, this will be Category:English footballers
2. Add a section of (drop-in replaceable) regex-based code to define the trial criteria. In the test case, look for a single main infobox (eg checked via Module:Is infobox in lead) based on {{Infobox football biography}}; and the first 30 words of the lead must include is|was a|an [+ 0 to 5 words] English [+ 0 to 5 words] football
3. The SD definition code as now, but with a dash (–) between years
4. Provide up-front information on the total number of articles to be processed (ie those within the selected category that have no SD)
5. Ignore for all purposes any article that already has a short description (or equivalently, the article is a member of Category:Articles with short description)
6. Add a user-assist mode flag that, when set, steps through the articles in the category one-by-one and presents to the user the draft SD along with enough information to check for errors: (1) the article, (2) the proposed SD, (3) any corresponding Wikidata description [not for use - just for sanity-checking], and (4) the first 80 characters of the lead.
7. Add a trial bot-mode flag that, when set, runs through the articles and writes the proposed SDs to a trial list on a Wiki page for user-review. As already coded, but excluding articles that already have a SD, and outputting in table or tabbed column format: (1) the article title as a wikilink, (2) the proposed SD, (3) any corresponding Wikidata description, and (4) the first 80 characters of the lead. In this mode, the script would not need to be logged in, and anyone could run it to try out various trial criteria and SD definitions.
8. Implement some form of user-control for printing of this trial list. The full list may be extremely long (hundreds of thousands of entries), and it may not be feasible to print everything out. The existing limit on number of lines is useful, but even better would be to spread the output across the whole of the input data. For example, if the user wants a sample of n articles out of a total of T with no SD, print every (T/n)th line. Also, Jonesey95 has asked for the ability to print the next n lines starting at a specified initial letter of the article title
9. Generate a separate exceptions output list, allowing the articles that fail the trial criteria to be checked manually. Same columns as the main list.
10. Add a run bot-mode flag for use once the trial criteria and SD definitions have been perfected for a particular category. In this mode, the script should write the new SDs at a bot-compliant speed directly to the relevant articles, keeping a log. Depending on the number of pages to be updated, this may need individual bot approval. That would be sought by whoever is proposing to do the run, based on the relevant trial lists after checking and approval. I now don't think we should bother with the read-back-the-trial-list option, as in most cases where a bot is required that list will be only a sample of the pages to be processed. MichaelMaggs (talk) 14:59, 14 August 2020 (UTC)

This is probably getting into too much detail for editors to comment on before seeing some actual outputs. But, Mike Peel, if you're still willing to implement this I'd be happy to do some test runs and work up a variety of specific scripting proposals that I could then seek consensus and bot approval for. MichaelMaggs (talk) 21:16, 19 August 2020 (UTC)

@MichaelMaggs: Mostly this seems OK. For performance reasons I think it would be best to only do two passes through the category, one to prepare the short descriptions and statistics, and the other to implement them. I've been trying to think of a good way to run this that keeps users in control without them having to run the python code and do bot approval themselves, but avoiding having to develop a complex user interface. My thoughts are currently:
1. Have an approved user list on a protected page (so we can control who has access), like:

• User:MichaelMaggs
• User:Mike Peel
• etc.

2. Have a run script page that the bot would keep an eye on (checking say once per hour) with a table like this one. We can define some standard formats like %BORNDIED%. Start/end could be the first letters of the articles, they might need to be more than the first letter if articles aren't ordered alphabetically. Going through subcategories could be an option, but would need a strict limit on how many levels to go into, perhaps just 1 level? And perhaps a maximum number to do at once (10k?).

Category	Username	SD definition	Must include	First sentence must include	Start	End	Maximum number	Holding page	Status
Category:English footballers	User:MichaelMaggs	English footballer (%BORNDIED%)	nfobox	English football	A	B	1000	User:Mike Peel/shortdesc	Ready/In progress/Done

3. If the editor is an approved one, then the bot could do part 1 to prepare the short descriptions and to post them to the holding page as a table like:

Article	New short description	Wikidata	Lead sentence
Ben Abbey	English footballer (b. 1978)	English footballer (born 1978)	Benjamin Charles Abbey (born 13 May 1978) is an English former footballer.

4. The editor could then change/remove these short descriptions as they want, and then add something like Category:Ready for Pi bot when they are done. (I would like to add an requirement that the modified descriptions would be CC-0 so they can be written back to Wikidata as well if there's not a description there already.)
5. Pi bot would then load the holding page and implement the entries under "New short description" (probably double-checking that the article is still in the category to avoid it being used to batch-set short descriptions for other article selections), then move the page to a category like Category:Done by Pi bot. Bot edits here are normally 1 every 10 seconds = 8640/day, so if fully loaded then we could get through all remaining articles in a year or so.
Perhaps this goes too far, particularly given that getting bot consensus in this area would be tricky. In which case I'm happy to go back to the original downloadable python script case. What do you think? Thanks. Mike Peel (talk) 17:06, 20 August 2020 (UTC)

@Mike Peel: That looks interesting and would certainly allow editors to contribute who don't have any Python experience and who don't want to get involved with bot requests. The practical issue, though, is that you'd have to create something pretty sophisticated to allow for the full flexibility I have in mind for the trial criteria and SD definition code snippets. They will need to vary quite a lot between runs, according to the article structure within each category, and it won't always be a case of finding specific words within the start of the lead, nor can it be assumed that the SD text will always be directly based on the category name. To some extent it will need trial and error, especially as the end result may well need consensus. It is that need for that flexibility that makes me think it would be easier for anyone who wants to contribute to edit and run a script and/or bot themselves (assuming you agree to your code being re-used in that way, of course). It would be possible if you prefer to do all the bot-runs yourself via Pi bot, but that would complicate things as you'd need to modify your own code each time and perhaps yourself seek bot approval for something new. If I or others could do that independently, it would be a lot less work for you. Fully agree with you on making the CC-0 issue clear, as I do hold out some hope that one fine day the Wikidata community will decide that these are suitable for use on Wikidata as well. MichaelMaggs (talk) 19:53, 20 August 2020 (UTC)

@MichaelMaggs: It would be good if we can talk about this by Skype/similar at some point soon. There is a trade-off between what a bot can do and what flexibility can be made available. I'm happy for others to reuse my code as they want (this is why I make it freely available), but I'm less willing to spend time modifying it for specific cases rather than coding something that works generally, and I think it's best to minimise the number of times that bot approval is needed. With "one fine day the Wikidata community" etc., the onus really is on the enwp community to be willing to collaborate with Wikidata, rather than continuing to reject opportunities to do so, sadly. Thanks. Mike Peel (talk) 20:24, 20 August 2020 (UTC)

@Mike Peel: good idea, I'll email you. MichaelMaggs (talk)

@Mike Peel: Hi Mike, did my email of last Satuday make it through to you? MichaelMaggs (talk) 20:05, 26 August 2020 (UTC)

@MichaelMaggs: Sorry for the slow reply, I've emailed you back now. Thanks. Mike Peel (talk) 10:16, 27 August 2020 (UTC)

Have just been reminded that I ought to get back to this project. MichaelMaggs (talk) 09:24, 18 November 2020 (UTC)

Acceptable Accuracy for fully-automated posting

If we want to post short descriptions using a script in a fully automated way (i.e. without a human manually checking each description), we need to decide at least roughly what an acceptable error rate looks like. It's worth noting that it is much harder to hit the error rate as it goes down. It's trivially easy to write something that makes an acceptable short description 90% of the time. 99% is hard, but might be doable with a lot of work on specific categories. 99.9% is going to be impossible for all but the most tightly formulaic articles. Error rate should be determined by a reasonable human looking at the short descriptions and making a judgement call.

For context, PearBOT, which operated on BLPs, had an error rate that seemed to be in the realm of 1%, and was approved. (none of the errors were BLP violations, they were grammar errors). My Species shortdescription generator has an error rate of about 5%. It generates much more descriptive short descriptions than Pear's (which were of the form [Nationality][Occupation]). It also skips far fewer articles than Pear's (Pear skipped 80%, I skipped 20%). I also put relatively little time into the algorithm, (~10 hours), and it's still pretty crude. I'd personally set the line for fully-automatic posting at 3%, but that's a judgement call, and I'd be interested in hearing numbers from other people below. Tazerdadog (talk) 03:04, 15 August 2020 (UTC)

Would there be a simple way to patrol such bot edits? · · · Peter Southwood ^(talk): 06:16, 15 August 2020 (UTC)

I imagine that if nothing else, a bot can and should keep a log. We also could have the bot generate the short descriptions before they go live, and have humans check samples to determine an error rate/common problems. Tazerdadog (talk) 05:49, 16 August 2020 (UTC)

Based on errors that are realistically trappable (ie not including cases where the article itself is factually wrong or has been wrongly categorised), 3% would be OK as an upper limit, though I'd be aiming at a lower figure. MichaelMaggs (talk) 07:54, 15 August 2020 (UTC)

It would be unreasonable to expect a better error rate from a bot than from humans. Do we have any idea of how often editors get it wrong? · · · Peter Southwood ^(talk): 17:44, 15 August 2020 (UTC)

I have made a proposal to improve consistency in the way dates are used within short descriptions. Please contribute here. MichaelMaggs (talk) 16:12, 30 November 2020 (UTC)

Whenever I click the random article button to find a page that may need a short description, there's a pretty decent chance I end up at one of the gazillion moth species articles that exist here (most recently Mesocolpia dexiphyma). The short description for all of these should be "Species of moth", in my view, and getting that to happen automatically would produce a huge bump in the progress bar of the project. Most of these pages have the template {{Speciesbox}}, so that would be the template to generate it from. It uses the scientific names, though, so deriving "moth" from it (or, more challengingly if we want to apply this more broadly, deriving common names from scientific names) might be a challenge. Thoughts? (please use {{ping|Sdkb}} on reply) {{u|Sdkb}} ^talk 01:16, 18 November 2020 (UTC)

Ah yes: the moths, the moths. The huge number of those were my inspiration for proposing a bot - see above at #Bot proposal to create short descriptions from scratch, for articles requiring one. The moths and indeed many species can't easily be done from {{Speciesbox}} alone, but are quite easy from a combination of infobox and category. You've reminded me that I must get back to that. MichaelMaggs (talk) 09:12, 18 November 2020 (UTC)

It's not too difficult to write a module that reads the content of a page and looks for text like "Category:Moth" then outputs a short description from that. Calling the module from {{Speciesbox}} would ensure that a short description exists. However, that route is resource-heavy and runs every time a page is loaded, so a bot that adds a short description to the wiki-text would be preferable. Perhaps give Mike Peel a nudge to see about using PiBot to do the moths as that requires no natural language processing. --RexxS (talk) 11:57, 18 November 2020 (UTC)

@Sdkb and RexxS: This should be trivial to do via the python code I've been writing with @MichaelMaggs:. Whether that runs or not depends on the enwiki community, though, I think Michael Maggs is going to put together a proposal about it soon (I haven't had much luck with bot requests here in the past!). Thanks. Mike Peel (talk) 19:36, 28 November 2020 (UTC)

Mike Peel, glad to hear! I stop by WP:BOTREQ on occasion, so I can offer my support there (with the caveat that my python skills are not great). {{u|Sdkb}} ^talk 19:40, 28 November 2020 (UTC)

That's great. I'm planning to put up a proposal for approval next week, and will suggest starting with the moths! MichaelMaggs (talk) 22:29, 28 November 2020 (UTC)

Hi User:Sdkb, you might like to have a look at the moth-related section below, and let me know what you think. MichaelMaggs (talk) 19:05, 7 December 2020 (UTC)

In this section above I suggested creating a bot to make serious headway in adding short descriptions to the several million articles that still don’t have one. User:Mike Peel kindly wrote some code to get me started with my first bot, and I’m now ready to go with ShortDescBot. The English footballers category was to be the initial task, but as someone has since completed that with AWB I’m proposing to start with the 26,000 or so moth articles that still lack a short description. The moths seem a good place to start since suitably-precise short descriptions can’t trivially be generated from existing inboxes (even for articles where one exists), at least without expensive Lua calls. No changes will be made to any existing short description.

The aim is to keep the new descriptions simple so that they can be added to many articles quickly, while still maintaining a low error rate. It’s always possible to hand-craft more sophisticated text, but with so many articles still to do we should I think be aiming for ‘decent’ rather than ‘perfect’ at this stage.

The procedure is, on a category-by-category basis:

1. Run the bot in trial mode, exporting all of the bot-proposed changes to a local spreadsheet
2. Review for obvious errors, adjust the code, and repeat 1 until the automated error rate is sufficiently low
3. Manually remove any remaining evident errors from the list
4. Run the bot in edit mode, making changes only to the pages in the final corrected list.

The moth articles are well structured, and it’s possible to identify “Species of moth” and “Genus of moths” with near 100% accuracy. You can see a sample of 200 or so proposed edits from Category:Moths of the United States at User:MichaelMaggs/Moths; note that the bot correctly identifies several articles as genus which Wikidata wrongly has as species. Of the 837 target articles in that category, the bot is able to fix over 98%, with just a few being skipped where it wasn't quite able to extract the first sentence of the lead.

The moths will be the first task for ShortDescBot. Assuming that’s successful I’ll then come back here to discuss future options, the next probably being a wider range of species and genus articles. WP:BAG approval will be needed for the task, and as BAG editors look for evidence of community consensus, it would be helpful to get some support in the comments below! Happy to take questions, too. Many thanks, MichaelMaggs (talk) 19:01, 7 December 2020 (UTC)

Discussion

• That's an impressive success rate! Is there any way to fix the incorrect Wikidata entries at the same time? (Or are automated solutions for that already in place?) —jameslucas ^{▄▄▄ ▄ ▄▄▄ ▄▄▄ ▄} 19:23, 7 December 2020 (UTC)
  • User:Mike Peel has Wikidata:Wikidata:Requests for permissions/Bot/Pi bot 14 which pulls new Wikipedia descriptions over to fill in blanks on Wikidata, but editors there are not keen on having their existing SDs automatically overwritten with the Wikipedia ones. MichaelMaggs (talk) 21:45, 7 December 2020 (UTC)
    • Keeping enwp and Wikidata descriptions in sync is a headache (and one I wish we'd avoided from the start). I'm thinking about writing a Wikidata Game that would ask whether the enwp or Wikidata description is better, but it could only save the change on Wikidata, which is a bit one-sided. As it stands, the best I can do is the automated pi bot script that imports descriptions to Wikidata where there's no description there. Thanks. Mike Peel (talk) 21:12, 9 December 2020 (UTC)
• Sounds alright to me; thanks for doing this! I wish it were a little more possible to tie the description to the infobox, so that if we decide to change something about the way we do short descriptions in the future it could just be changed in one edit there rather than having to run another bot task (the topic is on my mind because I recently came across a rather gnarly instance of the downstream effects of not templatizing). But, as you note, that doesn't currently seem possible, so this will certainly be an improvement over the status quo. {{u|Sdkb}} ^talk 19:27, 7 December 2020 (UTC)

Thanks for the comments. I've applied for bot approval for the task. MichaelMaggs (talk) 14:00, 14 December 2020 (UTC)

 You are invited to join the discussion at Template talk:Infobox song § Automatic short descriptions. {{u|Sdkb}} ^talk 07:11, 20 December 2020 (UTC)

I'd like to be able to go through any remaining Vital Articles that do not yet have a short description to add one. Is there any script that could be used for this purpose? (please ping me if you reply) {{u|Sdkb}} ^talk 19:34, 15 April 2020 (UTC)

Sdkb, I was going to suggest trying something with PetScan, but it seems to be down and, from what I can tell, has been for a while, and I'm not sure who if anyone is still maintaining it. Even then, I'm not sure where to go from there, because the "vital article" categories go on the talk page but the short description category goes on the article itself. Surely there has to be some sort of way to connect between the article and the talk page when making a query? Ionmars10 (talk) 20:35, 15 April 2020 (UTC)

Sdkb Ionmars10, PetScan is working now, how would we go about using it? comrade waddie96 (talk) 11:53, 20 May 2020 (UTC)

Waddie96, thanks for letting me know. I couldn't find a way to sort them by level, but here is a query that finds all vital articles that don't have native short descriptions. It takes advantage of the "templates&links" menu which, unlike the categories menu, has a "use talk pages instead" option. Ionmars10 (talk) 13:25, 20 May 2020 (UTC)

Thanks for the ping, Waddie96, and thanks for the tool, Ionmars10. Wow, 19,000 results out of 43,000 pages (44%) — that's more work than I'm prepared to take on. If we figure out a way to exclude level 5, it might be more manageable. In the meantime, if anyone does want to complete the level 5 list, I'd be happy to give them a barnstar; I'll add it to the WP:Rewards board. {{u|Sdkb}} ^talk 19:56, 20 May 2020 (UTC)

Sdkb, I just realized that VAs for each level are linked from Wikipedia:Vital articles and its subpages, so here are queries for level 3 and level 4. As of writing, these return 53 and 1981 results respectively. Ionmars10 (talk) 20:52, 21 June 2020 (UTC)

Ionmars10, oh nice! That's much more manageable. {{u|Sdkb}} ^talk 20:54, 21 June 2020 (UTC)

Update: completed level 3! {{u|Sdkb}} ^talk 20:13, 22 June 2020 (UTC)

Sdkb, that would certainly be a useful script to have to iron out any vital articles that do not have a local short description. I mean it was only recently that I added a short description for the vital level 3 article Michelangelo [1]. Regards  Spy-cicle💥  ^Talk? 21:59, 15 April 2020 (UTC)

• Hello, Sdkb and Ionmars10, the petscan link on the reward board is yielding no results. Is that because there are no more short descriptions to do on level 5 vital articles, or because the petscan is still configured to scan a vital article level that has already been completed such as 3 or 4? I tried various search queries, and still yielded no results, but I'm not certain the exact search strings I should be using. Thanks. Huggums537 (talk) 05:21, 16 January 2021 (UTC)

  Huggums537, are you using the very first Petscan link in that section of the checklist? Just speculating, but it might be that the query is timing out because of how long it takes to load. Level 4 still seems to work fine and currently returns 914 results, but the level 5 one is notorious for how long it takes to load due to the sheer number of articles. Ionmars10 (talk) 05:50, 16 January 2021 (UTC)

  It appears to be the only link in the section Wikipedia:Reward_board#Completing_short_descriptions_for_Level-5_Vital_Articles. I'm not sure what checklist you are talking about unless it's this one: Wikipedia:WikiProject_Short_descriptions/Progress#Vital_articles, and then, yes I tried that first link too, and it took a long time to load just like the link on the reward board, but yielded no results. Huggums537 (talk) 06:03, 16 January 2021 (UTC)

  Huggums537, yes that's the one I was talking about. Sorry, to be honest I completely forgot that the reward board was even a thing. Ionmars10 (talk) 15:11, 16 January 2021 (UTC)

• Maybe try the level 5 search for a single subject at a time so that it does not overload. This one is just the mathematics articles and returns 312 results — GhostInTheMachine ^{talk to me} 11:50, 16 January 2021 (UTC)

  Thank you, GhostInTheMachine. That should assist me in searching specific categories. I appreciate your help! Huggums537 (talk) 15:38, 16 January 2021 (UTC)

How do I find a list of all articles without a short description?

Search for -incategory:"Articles with short description" — GhostInTheMachine ^{talk to me} 23:02, 5 January 2021 (UTC)

@GhostInTheMachine: - This doesn't appear to work, I conducted this search entry and I only found articles with a short description. Are you sure this is the right query? Thanks :) - UnguidedEmperor

Did you include the minus sign in front of incategory? — GhostInTheMachine ^{talk to me} 23:53, 5 January 2021 (UTC)

@GhostInTheMachine: Yep. - UnguidedEmperor

should be about 2.5 million without the minus sign (has a SD) and over 3.6 million with the minus sign (not in the cat and so without a SD). Try this — GhostInTheMachine ^{talk to me} 00:03, 6 January 2021 (UTC)

@GhostInTheMachine: - Even using your link, I am still only seeing articles with a short description. It is listed at 3.6 million with that specific link as well, if that helps. Most likely something on my end. - UnguidedEmperor

do the search that returns 3.6 million results and give me the first article title00:18, 6 January 2021 (UTC)

@GhostInTheMachine: Done, at this moment it shows Driveshaft, however this changes once I refresh the page, on 2-3 refreshes it now shows Point guard. - UnguidedEmperor

Skip Driveshaft – it is a redirect. Point guard is OK, it does not have a SD. Check the cats or the raw text. If you are using the helper then it will be showing you the description currently in Wikidata and offering you the chance to import it – which creates the SD. — GhostInTheMachine ^{talk to me} 00:27, 6 January 2021 (UTC)

@GhostInTheMachine: - Oh, that explains it, sorry. So in what cases should I import the short-description from Wikidata? Or maybe perhaps you could link me somewhere explaining this. Thanks for the help again. - UnguidedEmperor

Just jumping in here. If you're using the gadget you can accept the Wikidata suggestion provided that it complies with the Wikipedia guidelines at WP:HOWTOSD (but don't worry if the Wikidata text starts with a lower case character, as that will be corrected to upper case when you accept it). Depending on your fields of interest you may find that some or even most Wikidata suggestions can be accepted without amendement. Many, though, can be very quickly and easily improved. MichaelMaggs (talk) 14:03, 18 January 2021 (UTC)

I just noticed Loyola Academy is in Category:Articles with long short description even though its SD is short. I don't see any recent changes to the SD. This article has a infobox-generated long SD that has been overridden with a local SD. I'm guessing the category is added by the infobox template even though the local SD is the one that "counts". MB 15:53, 30 January 2021 (UTC)

The article has two short descriptions:
(1) A default SD set by the infobox from the School Type and other parameters — Private, catholic, non-profit coeducational secondary (grades 9-12) education institution school in Laramie Avenue Wilmette, Illinois, United States — 148 characters
(2) An explicit one set by the short description template at the top of the article — Catholic college prep school in Illinois, U.S. — 46 characters
The explicit SD overrides the infobox SD for display purposes, but the infobox SD is setting the category — GhostInTheMachine ^{talk to me} 18:30, 30 January 2021 (UTC)

Right, but isn't the intention of the category to provide a list of articles for which the "long" SD should be fixed. An article that has a "compliant" SD shouldn't be in the category. MB 18:51, 30 January 2021 (UTC)

@MB and GhostInTheMachine: I've created a template Template:Has short description that returns the position of the phrase {{short description within the article wiki-text, or nothing if not found.These are typical results from the template

• {{Has short description |title=Loyola Academy}} → 1
• {{Has short description |title=Rock Bridge High School}} →

I've now added it to Template:Infobox school/short description/sandbox to suppress the output if there is already an explicit short description on the page.

I've tested it at Loyola Academy, which is no longer in Category:Articles with long short description, and at Rock Bridge High School, which still generates a short description from the infobox.

It will probably want some further testing before updating the main template Template:Infobox school/short description from its sandbox. Cheers --RexxS (talk) 20:57, 30 January 2021 (UTC)

{{#invoke:string2|findpagetext}} – Dang!! There is always something more to learn. I should have used it for the short description not first cats. I will try it in the SD sandbox — GhostInTheMachine ^{talk to me} 22:54, 30 January 2021 (UTC)

You probably missed it because I only wrote it on 9 December 2020. --RexxS (talk) 01:26, 31 January 2021 (UTC)

The following two articles came up in my petscan search results as not having descriptions Concordat of Worms and Public_international_law, but one redirects to a specific "anchor" section in the article, so I checked the main article descriptions as well as the redirect pages, and all of them had descriptions, so why did these two items show up in the search results as having no descriptions? How do we solve this? Huggums537 (talk) 11:23, 5 February 2021 (UTC)

@Huggums537: I don't think there's anything to solve with the short descriptions.

If I start to search using a mobile view for "Public international law", by the time I get to "public intern", I am offered "International law Generally accepted rules, norms and standards in international relations as one of several options. By the time I get to "public internatio", I'm down to two options: Public International Law & Policy Group and International law.

If I start to search using a mobile view for "Concordat of Worms", by the time I get to "concordat", I am offered "Investiture Controversy 11th- and 12th-century dispute between secular rulers and the papacy as one of several options. By the time I get to "concordat of worm", I'm down to the single option: Investiture Controversy.

I believe that is working exactly as intended: searching for a redirect displays the short description for its target in the places where a short description is useful. Both redirect pages, Public international law and Concordat of Worms, have short descriptions of "Redirect page", which do precisely nothing. I think any problem may lie with your petscan search. --RexxS (talk) 15:44, 5 February 2021 (UTC)

I don't really know how to create my own petscan search queries from scratch, so I modified this scan by Ionmars10 and combined it with this scan by GhostInTheMachine to come up with my own. The articles in question should be the first two results, but all the rest don't have any descriptions as they are supposed to... Huggums537 (talk) 18:09, 5 February 2021 (UTC)

@Huggums537: the problem is that your PetSCan search excluded Category:Articles with short description, but Concordat of Worms and Public international law aren't in that category; they are in Category:Redirects with short description. I've added that category to be excluded to your PetScan query at https://petscan.wmflabs.org/?psid=18379015 and the number of results has reduced from 3825 to 3823. Cheers --RexxS (talk) 21:54, 5 February 2021 (UTC)

Wow, I was worried for nothing. I was thinking there would be way more "problems" to deal with, and I thought there would be others besides the two I already found, but it appears I "got lucky" and accidentally found the only two there were. Well, at least I still have a huge pool of descriptions to work on. Thanks. Huggums537 (talk) 22:43, 5 February 2021 (UTC)

I have seen a small number of articles where the SD template has a leading space {{ short description... – and in one case 2 spaces. In theory this is OK for the template itself, but the Shortdesc helper gadget does not recognise this as a SD template. In a couple of cases, this led to somebody adding a second SD via the gadget. I have fixed all cases of this, added it to my regular checks and posted a change request. — GhostInTheMachine ^{talk to me} 15:17, 6 February 2021 (UTC)

It seems that the SD on list articles are all set to "Wikimedia list article" in WD. Any guidance on this? The title is usually self-explanatory (e.g. List of hotels in Chicago). Should we just import this, change it to "Wikipedia list article", or something else? MB 04:58, 28 April 2020 (UTC)

It may be worth reviewing the section above titled wikimedia project page. I think the current consensus is to improve it if you can, otherwise use "Wikipedia list article". --RexxS (talk) 16:39, 28 April 2020 (UTC)

Now that we have the option of "none" where no SD is needed, are we doing to change all the "Wikimedia list article" descriptions to that, or leave them as they are indefinitely? MichaelMaggs (talk) 12:15, 10 February 2021 (UTC)

ShortDescBot task 2 is ready to start work on organism articles, as discussed above, and is awaiting approval for the task. Where the bot writes a short description for an article on a genus which is monotypic (that is, where the genus contains only a single species), it currently phrases it (for example) "Single-species genus of butterflies". I chose that in preference to "Monotypic genus of butterflies" as the word "monotypic" is I suspect largely unknown except to biologists, and doesn't comply with WP:HOWTOSD which recommends "simple, readily comprehensible terms that do not require pre-existing detailed knowledge of the subject." "Single-species" is, I believe, both comprehensible to the majority of readers and is also accurate.

But on Wikipedia:Bots/Requests for approval/ShortDescBot 2 Mr Fink, who has some specialist knowledge, has indicated his preference for 'Monotypic'. I have no problem switching to that if there is consensus to do so, but would like to know what others think of the importance or otherwise of sticking to comprehensible terms. If there's no consensus, or no particular interest, I'll simply drop the idea of flagging that type of genus article, and stick with "Genus of butterflies". MichaelMaggs (talk) 17:20, 7 February 2021 (UTC)

That works, too. Worse comes to worst, someone else can come swoop in to further modify the short description.--Mr Fink (talk) 03:05, 8 February 2021 (UTC)

I made a comment elsewhere also, personally I favor monotypic for two reasons, first it may have originated in jargon but is widely used these days, not just to biologists but also in high-school textbooks and the media etc. Also second what will you do with monotypic families (single-genus?), monotypic Orders (single-family?) it seems to me it starts to get absurd, when there is a well known word. When I give lectures on Biology its not always to my peers, in general lectures at schools or public audiences I use the term monotypic and have never been asked what it means, people know this word. Cheers Scott Thomson (Faendalimas) ^talk 04:32, 8 February 2021 (UTC)

I'm not sure what circles you move in, but I seriously doubt that the average reader has any clue what "monotypic" means. And we write for the average reader, not for specialists. I taught in secondary schools for 25 years and I'm absolutely certain that not a single pupil I ever taught would know what "monotypic" meant. Have you considered the possibility that your lecture attendees were simply being polite when they didn't understand?

We're not talking about families or orders, just genera, although I accept that "single-species genus" relies on the reader understanding "genus", so maybe it's impossible to write a jargon-free short description in some cases. --RexxS (talk) 05:33, 8 February 2021 (UTC)

20 years ago I would agree with you, world has changed. No I can gather they know the meaning of a word when they use it in their questions correctly, and not to ask what it means. Alot of these words originate in jargon, in the original set up of the concept of genera and species (1750s) all life forms were a genus (form) which were identified by specifica, which became species which was a description of the genus, the specifica could be up to 6 words. Obviously that has changed. Language evolves. Cheers Scott Thomson (Faendalimas) ^talk 05:55, 8 February 2021 (UTC)

You may have a point. I left teaching in secondary schools in 1997/8, so it's conceivable that standards have risen monumentally since then. Although I usually consider 20 years ago as "relatively recent". --RexxS (talk) 06:25, 8 February 2021 (UTC)

Change in the age of the internet is fast, in 1980 when I finished highschool I did not know what monotypic meant. Yes I spend most of my time in Universities, preaching to the converted. But I do lecture and provide tuition to secondary students and give seminars, etc to the general public. I know I have to change my language for a given audience, it is rude not to. I have done this in 3 countries. This is not one of the words I need to change these days. Anyway not trying to make an argument here, I was asked to comment and the decision will take all comments into account, in the end I will be fine with the decision whatever it is. Cheers Scott Thomson (Faendalimas) ^talk 06:41, 8 February 2021 (UTC)

I'm requesting bot approval for "Single-species", as that's the only option that's widely comprehensible to non-specialists when seen without context, as it will be, embedded within a mobile phone search list of article titles. While I do understand that specialists may prefer their own specialist term, strongly arguing for that here is likely to be counterproductive. Lack of consensus will mean I can't do either, and tens of thousands of genus articles will have a less informative short description than I could easily provide. MichaelMaggs (talk) 09:45, 8 February 2021 (UTC)

Single species (14 char) is slightly more user-friendly, monotypic (9 char) is slightly more compact. I would not bother to change one to the other as both work adequately. You could go with whatever is used more often in short descriptions to date on the principle of most common usage on Wikipedia. If this is to fill in short descriptions where they do not yet exist, both options are better than nothing and should not require a formal consensus as either is an improvement, and can be changed later if there is any good reason to do so. · · · Peter Southwood ^(talk): 12:49, 10 February 2021 (UTC)

Thanks for the comment. This is for new SDs only. The reason it needs some consensus is that the question has been raised by a member of WP:BAG as part of the bot approval process. MichaelMaggs (talk) 13:00, 10 February 2021 (UTC)

Unless you have good reason to do otherwise, it's best to use the same convention that is used on Wikidata, as otherwise you will generate a lot of discrepant descriptions for no reason. Thanks. Mike Peel (talk) 21:15, 8 February 2021 (UTC)

Well, the reason would be to ensure they are compliant with WP:HOWTOSD and MOS:JARGON, which apply here but not on Wikidata. MichaelMaggs (talk) 12:25, 10 February 2021 (UTC)

This could use some more opinions. Pinging editors who’ve shown an interest, from this section. @Jonesey95, Dyanega, GhostInTheMachine, Pbsouthwood, and Certes: It's this very specific question I need feedback on: in the context of a short description for a genus, should the bot use “Single-species” or “Monotypic” for the relevant articles? So far I see no consensus which means I won't be using either. MichaelMaggs (talk) 12:04, 10 February 2021 (UTC)

Either option seems good to me. The objective is to distinguish the genus from (say) the Greek city and the rapper with similar names. Unless a plant and animal share the name, "genus" is normally enough and any further information is a bonus. As the description will be well within the 40-character advisory limit, I would go for the extra clarity of "single-species". Monotypic is less widely understood, especially by younger readers and non-native English speakers. (I am neither a biologist nor a prolific adder of SDs.) Certes (talk) 13:11, 10 February 2021 (UTC)

I would stay away from "monotypic" in a short description. I recommend "single-species" if you must, or just "genus of butterflies", which should be perfectly fine to distinguish the genus from things like album names, bands, and sculptures. Remember that the primary use of SDs is to distinguish among articles with similar names; once you have accomplished that, you are usually done. – Jonesey95 (talk) 15:25, 10 February 2021 (UTC)

I see three problems, one of which has been covered by most of the preceding comments. Another is one very briefly raised, namely that any taxon group from genus upwards can be monotypic, so you can't just have the bot replace the word "monotypic" with "single-species" because there are monotypic genera, monotypic families, and potentially other ranks. This increases the difficulty of replacing things globally without either messing something up, or overlooking something (there might be a small number of pages described as monotypic subgenera or monotypic superfamilies, for example). Finally, as noted in earlier threads but not yet in this one, CLASSIFICATIONS CHANGE, sometimes incessantly, as is very common in insects. By adding in EITHER type of qualifying term, be it "monotypic", "single-species", or any such thing - heck, even modifiers like "large genus" or "small genus" which seem pretty harmless - you run the very real risk of the SD and the article it describes becoming disconnected from one another if the constituency of a group changes. I would argue fairly strongly that NO such modifiers should be allowed in SDs. Even as it is, there are enough changes in rank happening every day that just the bare-bones SDs saying "family of X" or "genus of Y" are prone to becoming outdated, and justify keeping taxonomy-related SDs short and very simple (modifier-free, and not using taxonomic names internally). The amount of work required to keep things coordinated should be as small as possible, and these modifiers are just one more thing to keep track of. Dyanega (talk) 17:14, 10 February 2021 (UTC)

I am inclined to agree with @Dyanega: here. Earlier I was just looking at this from an a or b scenario ie monotypic or single-species, and I for saw problems with single-species, but in reality for what your trying to achieve here why does it matter if the genus has one or 1000 species in it for purposes of identifying more or less what it is. How many species are in the genus is extended information best achieved by visiting the page, all the reader needs is to figure out what the genus is in terms they will understand. I suggest to drop this term, either of them, altogether. Cheers Scott Thomson (Faendalimas) ^talk 17:37, 10 February 2021 (UTC)

• Genus of butterflies is fine. Is Genus of butterfly more correct? Perhaps Genus of carnivorous butterfly, if that is one of their distinguishing features. Single-species ... (or Monotypic ...) does not really help the reader if they were looking for a rock band — GhostInTheMachine ^{talk to me} 17:23, 10 February 2021 (UTC)

Thanks all, those comments are really useful. I find the arguments against using both "Single-species" and "Monotypic" persuasive, and I'll change the bot so that it doesn't make use of either. MichaelMaggs (talk) 09:34, 12 February 2021 (UTC)

I think that's the right decision. Although I didn't argue for it, I should have done. Few readers will select a genus because it is monotypic but its homonyms aren't. Even in a case like Burkillia, the information that it is an alga rather than a snail or legume is far more useful than mentioning that it has only one species. Certes (talk) 11:14, 12 February 2021 (UTC)

That does avoid the problem. If anyone feels strongly enough that single-species or monotypic is worth adding in any given case, they are free to do so. In the meanwhile, we get a large number of adequate short description from the bot. · · · Peter Southwood ^(talk): 13:56, 12 February 2021 (UTC)

What are we doing now for articles that don't really need a SD. For example, I don't see that a SD for Alpine skiing at the 1960 Winter Olympics – Men's downhill would be helpful in anyway. I recall something that acknowledged there are cases like this. Wikipedia:Short description says "Most mainspace articles should have a short description." but does not mention exceptions. Should this article have {{Short description|none}}? MB 16:20, 7 February 2021 (UTC)

@MB: yes, and the documentation should be updated to indicate that we can now have blank SDs.--RexxS (talk) 18:37, 7 February 2021 (UTC)

I added a section in Wikipedia:Short description. MB 19:19, 7 February 2021 (UTC)

I take it we're going with |none rather than |None then. That's OK. MichaelMaggs (talk) 19:57, 7 February 2021 (UTC)

If this is to add short descriptions to articles which don't have one and don't need one, then "none" will be fine. If there is already a short description manually added, don't override by bot unless it means the same as "none". If you think "none" is better than what is non-trivially already there, make a manual change and say why. Otherwise there will probably be complaints. Cheers, · · · Peter Southwood ^(talk): 12:33, 10 February 2021 (UTC)

What about the SD helper tool. Should it display something to indicate no SD has been explicitly selected? On Desiigner discography, it offers to import from WD while on Alpine skiing at the 1960 Winter Olympics – Men's downhill is says "missing". In both cases, there is a local {{Short description|none}}. MB 16:56, 17 February 2021 (UTC)

That's odd. In 2015 in Bangladesh it displays "This page has deliberately no description". In the first two cases (described above), the Local description field in the Page information is nil and there is no SD shown in a mobile search. For the third, the mobile search first displays "List of events" and then "Calendar year" as more characters are added. The caching makes it impossible to test properly. --RexxS (talk) 18:18, 17 February 2021 (UTC)

What is the status of {{short description|none}}? The gadget does not see this as an article having a valid SD and offers to import a SD from Wikidata — GhostInTheMachine ^{talk to me} 15:25, 6 February 2021 (UTC)

@GhostInTheMachine: there are around 358 articles using {{short description|none}}. Search on hastemplate:"short description" insource:"{{short description|none}}".

List of countries by Human Development Index had two short descriptions: {{short description|Wikipedia list article}} and {{short description|none}}, so I removed the 'Wikipedia list article' one. Then the Page information showed the Central (i.e. Wikidata) description as 'Wikimedia list article', but the Local description had disappeared. Doing a mobile search for list of countries by h offered List of countries by Human Development Index with no short description.

Next I removed the |none leaving just {{short description}}. That gave the same result for the Page information and the mobile search, showing that a blank SD and SD=none are equivalent.

That means we can now have blank short descriptions, or use {{short description|none}} to indicate it is intended to be blank, with the same result. We are now in a position to remove {{short description|Wikipedia list article}} (or Wikimedia list article once the SD helper is fixed to accept blank SDs or SD=none (so that eager editors don't continually re-add the redundant descriptions from Wikidata). --RexxS (talk) 19:08, 6 February 2021 (UTC)

We should specify the use of SD=none. A bare SD template looks like somebody just forgot to add the description. Adding none looks deliberate. — GhostInTheMachine ^{talk to me} 19:31, 6 February 2021 (UTC)

Indeed, and that's why I originally coded it that way. --RexxS (talk) 23:10, 6 February 2021 (UTC)

It seems there's agreement about this, but it's unclear to me who, if anybody, needs to make the necessary changes to ensure that it works properly. Am I right in understanding that RexxS's code needs no modifications but that the gadget does? And while we're discussing it, wouldn't {{short description|None}} be better than {{short description|none}}, given the guidance that SDs should start with a capital letter? I'm happy to update the instructions in various places once I know what's been agreed. See also the query below. MichaelMaggs (talk) 16:46, 7 February 2021 (UTC)

If you want the SD to explicitly indicate a blank value by using |none, then the template needs no modification.

If you want the SD to explicitly indicate a blank value by using |None, then change the template and change the 358 articles that use "none".

If you want to avoid the SD helper missing the explicitly blank short description and offering to add another SD template, then the SD helper code needs modification. --RexxS (talk) 18:34, 7 February 2021 (UTC)

The template should use a case insensitive test for "NONE" etc. — GhostInTheMachine ^{talk to me} 22:52, 7 February 2021 (UTC)

I've implemented a case-insensitive test in the template. I haven't checked whether the SD helper does a case-insenstive test. --RexxS (talk) 00:02, 8 February 2021 (UTC)

I think that's a relic from when "none" wouldn't do anything since it wouldn't override the Wikidata description. I'll fix the gadget so it accepts empty descriptions. Galobtter (pingó mió) 18:39, 7 February 2021 (UTC)

• The SD helper is accepting a description of "none", but then indicates that the SD has not been set at all, but should still be imported from WikiData — GhostInTheMachine ^{talk to me} 20:33, 23 February 2021 (UTC)

Should we change the guidance to reflect a new way of thinking about improving descriptions now that we have met our "quota" and the wikidata has been turned off for mobile?

Follow the discussion here: Wikipedia_talk:Short_description#"Duplicate_information" Huggums537 (talk) 10:37, 26 January 2021 (UTC)

 Comment: According to WP:RFCNEUTRAL RFCs aren't supposed to be for questions like "What do other editors think about the discussions on this page?", which is all this boils down to. MichaelMaggs (talk) 12:08, 26 January 2021 (UTC)

MichaelMaggs, Thanks for the link to WP:RFCNEUTRAL. I didn't realize the question was inappropriate. It did seem innocent enough at the time. Huggums537 (talk) 12:40, 26 January 2021 (UTC)

This discussion was resolved. RfC template removed. Huggums537 (talk) 12:22, 26 January 2021 (UTC)

The red/green state of the project bar shows the project as currently 47.797% complete, but I think that may be a significant underestimate. If I understand the calculation correctly, the red part of the bar includes all articles (except disambiguations), and in particular it includes redirects. Since very few if any redirects need a short description, they ought to be excluded as well. Is there a way with a magic word or lua module to count how many there are and remove them? As there are so many it could make a big difference to our progress indicator. MichaelMaggs (talk) 22:26, 3 March 2021 (UTC)

According to Wikipedia:Database reports/Page count by namespace, we have 9.4 million mainspace redirects; articles plus dabs total 6.2 million. According to Petscan, 3554 mainspace redirects have short descriptions. So redirects can't be included in the divisor for that SD percentage. Certes (talk) 23:29, 3 March 2021 (UTC)

Going off at a tangent: should redirects ever have short descriptions? Should they be removed either selectively or wholesale? Should Shortdesc helper advise editors to go away when they try to add them? Certes (talk) 00:28, 4 March 2021 (UTC)

You could argue the existing descriptions are not doing much harm, but I'd prefer to get rid of them and to simplify the guidance to clarify that SDs should not be added to redirects. We currently make redirects an exception to the general rule that SDs should go at the top of the page, which seems pretty pointless when we actually don't want them there at all. ShortDescBot could remove existing SDs from redirects if people think that's worthwhile. MichaelMaggs (talk) 04:44, 8 March 2021 (UTC)

I can see a case for some, e.g. Willow Park School, Dublin is correctly identified as a school rather than the entire settlement covered by the target article. Others are harmless, such as 1590 in Belgium: both the topic being redirected and the target are lists. I was thinking more of Bishop of Kutigi (Anglican) whose rather meta SD explains that the page is a redirect. We might need a bot to find those, as I can't think of a way to search the wikitext of a redirect rather than its target. Certes (talk) 14:30, 8 March 2021 (UTC)

Fairly easy to do with a bot, as Pywikibot accesses redirects without following through to the target page. As long as there is something specific that needs to be searched for. MichaelMaggs (talk) 14:43, 8 March 2021 (UTC)

In the case of redirects with possibilities a short description can indicate the scope of the potential article, which I think is useful. There may well be cases where a short description is not useful for a redirect, but I cannot think of an obvious simple rule for identifying them. If there are cases where a short description is harmful in any way, I have not noticed them. If not broken there is no need to fix. · · · Peter Southwood ^(talk): 18:58, 8 March 2021 (UTC)

At 06:10 UTC on 18 March 2021, the English Wikipedia officially crossed the 50 percent threshold on short descriptions, with 2,975,796 written descriptions! Sammi Brie (she/her • t • c) 06:11, 18 March 2021 (UTC)

I was just about to post the same. Congratulations to all! (I wonder who the lucky editor was, although I imagine we'll never know.) Extraordinary Writ (talk) 06:21, 18 March 2021 (UTC)

Nice! Here's to another 50%. ExcellentWheatFarmer (talk) 21:54, 18 March 2021 (UTC)

🎉🎉🎉🎉🎉 //Lollipoplollipoplollipop::talk 06:30, 23 March 2021 (UTC)

ShortDescBot has successfully completed its addition of new short descriptions to all the moth articles. Next, I want to move on to categories of other organisms. This is a good bot task since non-technical short descriptions complying with WP:HOWTOSD can’t automatically be generated from the usual infoboxes, at least without expensive Lua calls. There are around 210,000 target articles.

Each bot run is based on a single category at some level in the tree that I can manually associate with a suitable common generic name. Sometimes that may be the same as the category name (Category:Butterflies --> "butterfly"), but often not (Category:Poaceae --> "grass" or Category:Onychophorans --> "velvet worm"). The bot then constructs and adds new short descriptions such as "Species of butterfly", "Genus of velvet worms", "Family of grasses" and so on. The text is deliberately simple so that a low error rate (<1%) can be maintained while minimising the number of non-standard articles that the bot has to skip as 'too difficult to parse'. For each category the procedure is:

1. With the bot in trial mode, write the proposed descriptions to a local spreadsheet; review and repeat until the error rate is sufficiently low
2. Manually remove from the list any obvious classes of article that the bot will not realistically be able to handle [not had to do this so far in testing]
3. Re-run the bot in edit mode, making live changes only to the pages in the final corrected list.

The bot won't change existing short descriptions, with one small exception. A new feature this time is the inclusion of "Extinct ..." in the bot-created description of extinct organism articles, and also "Single-species .." in Monotypic genus articles (where that can be done without making the text too long). 1300 2000 or so of the moth short descriptions that the bot created last time can be improved.

You can see a sample of suggested edits from a variety of categories at User:MichaelMaggs/ShortDesc.

Feedback would be useful, please, before I apply for bot approval. MichaelMaggs (talk) 13:52, 18 January 2021 (UTC)

Discussion

I've looked at the source code, but I'm not sure I'm understanding it correctly. You want to: a) identify that an article is about an organism; b) determine the rank (genus/species/etc.); c) determine a generic common name; d) determine monotypy or extinction if applicable. I gather a) is accomplished by presence of one of the taxobox template, b) and d) are parsed from the lead sentence, and c) comes from membership in a particular category. Am I understanding correctly? Plantdrew (talk) 23:55, 19 January 2021 (UTC)

Plantdrew, : Generally yes, with some additional cross-checks (for example species articles aren't expected to have a single-word title, and identification doesn't rely on the wording of the lead as that can be very variable). The generic common name isn't bot-determined – and indeed in a large number of cases that's not possible – but is chosen manually as as suitable descriptor for the organisms within the category. In tests I've been finding that the approach works very reliably and allows a much more granular approach than was used for the Wikidata short descriptions, which mostly lump all plants together under the common name "plant". MichaelMaggs (talk) 05:49, 20 January 2021 (UTC)

There are a lot of inconsistencies in terms of what organisms articles "should" have and what they actually have. A species article could have a single word title (if it's a common name). Categories for monotypy are often absent (and articles on fish mostly use the binomial title for monotypic genera contrary to MONOTYPICFAUNA). Categories for taxonomic ranks (e.g. Category:Gastropod genera) are often massively underpopulated. I suppose cross-checks can bridge the gap with some (many?) of the inconsistencies.

Increased granularity is good, but I wonder how achievable it is. For a very large number plants, I don't think there is any good term that is more granular than "plant" (or "flowering plant"). Perhaps "tree" (when applicable)? But tree would either depend on a very underpopulated category again, or parsing the lead. I wish you the best in figuring out how to increase granularity.

Consider having "prehistoric" or "fossil" in the short description rather than "extinct". Wikipedia has separate categories for recently extinct organisms and those known only from fossils (although some articles are miscategorized). There are far more articles on prehistoric organisms than recently extinct organisms. Plantdrew (talk) 02:28, 27 January 2021 (UTC)

Plantdrew, That's helpful, thanks. On your first point, species that have single-word common name titles, such as platypus, are fortunately pretty rare, and as their leads are typically quite discursive and non-standard the bot can skip them and add them to its list of articles to be handled manually. Monotypy is often identifiable from the lead, even where proper monotypy categories are lacking, but it is quite common for the article to be entirely unclear. But where it is, the default description "Genus of ..." should still be correct. I wasn't aware of the problem of fish genera, and when I get to that I'll make some small changes to the bot's logic.

I'll need to experiment to see how much granularity is achievable, as the level within the tree that can be associated with a common name varies widely. For plants, categories that can be labelled "conifer" or "grass" are straightforward, but I'm sure I'll end up have to sweep up many categories into some high-level name such as "flowering plant".

I did notice there have been some efforts to distinguish between fossil and recently-extinct organisms, but those are inconsistent and incomplete and I'm not able to accurately separate the two. Rather than attempting the word 'fossil' and finding it applied to something that's only recently gone extinct, I've preferred to use 'extinct' in all cases. That should almost always be correct.

I don't suppose you know of a tool for visualising selected parts of the Wikipedia category tree do you? If such a thing exists it would be invaluable for helping to work through systematically. MichaelMaggs (talk) 09:55, 27 January 2021 (UTC)

The only category tree visualisation gizmo that I know of is the categorytree tag — GhostInTheMachine ^{talk to me} 10:50, 27 January 2021 (UTC)

Thanks, I didn't know about that special page. MichaelMaggs (talk) 14:27, 27 January 2021 (UTC)

 Done All done, so far as I can tell, but the category structure is so complicated that I may well have missed some categories. Feel free to let me know on my talk page if you come across any missing organism categories that are amenable to bot-generated descriptions and which are too large to handle manually. MichaelMaggs (talk) 17:31, 8 April 2021 (UTC)

Not sure how many users are still active around here, but I'd just like to draw attention to this category Category:Articles with long short description. There's many thousands of articles with too long short descriptions, which will need to be fixed at some point. This also can't be done by any bot, so any effort that you can give would be great. I'll try and give it a go when I have time. Regards, Willbb234^Talk (please {{ping}} me in replies) 23:45, 27 February 2021 (UTC)

Quite a few of these are the result of an infobox generating a description based on the box's parameters regardless of their length. For example, the description generated by {{Infobox school}} is usually something like [school type] in [place]. In the case of Assumption Samutprakarn School that results in Private roman catholic non-profit all-boys basic (primary and secondary) education institution school in Samut Prakan, Thailand. Overriding all of these manually will be a significant task. I'd suggest modifying the infobox code to return something shorter such as "School in [country]". Simpler and shorter text that always complies with WP:HOWTOSD is better than something complicated that often has to be fixed manually later. MichaelMaggs (talk) 15:48, 28 February 2021 (UTC)

Pinging Trialpears who I think wrote most of the SD code for that particular infobox. Would you be amenable to simplification? And maybe looking at other boxes that have the same issue? MichaelMaggs (talk) 22:41, 3 March 2021 (UTC)

MichaelMaggs, Yep, there are certainly some issues here that wasn't detected in part because the tracking didn't exist yet. I do not think removing the logic entirely would be appropriate since in most cases it is an improvement and significant discussion and testing went in to the current version. My proposed solution would be using the current logic, check if it generates too long of a description and if it does just use School instead of the modified type parameter it currently uses. --Trialpears (talk) 00:03, 7 March 2021 (UTC)

Trialpears, that sounds ideal. MichaelMaggs (talk) 04:32, 8 March 2021 (UTC)

MichaelMaggs, sorry I kind of forgot about this. I've just implemented this with a cut-off at 40 characters. Possibly a bit high but should remove most things from the cat. Sadly it doesn't fix when people put "Landhuis Street / Jim Fouche Drive, Roodepoort, Johannesburg" as the city. The location should be properly prioritized to be concise when used properly though. --Trialpears (talk) 23:21, 26 March 2021 (UTC)

Trialpears, MichaelMaggs I attempted to change the SD of I SS Panzer Corps but the original SD remained and I had to go into the wikitext to remove it. This is likely an error on the part of the Shortdesc helper but it could be to do with the SD being too long or was formatted incorrectly (the diff). Another point is the category mentioned above seems to have some false-positives in it. See 460th Space Wing, Lock and Dam No. 25, and Lock and Dam No. 22 which all have short SDs but still appear in the category. Kind regards, ^Talk (please {{ping}} me in replies) 22:51, 26 March 2021 (UTC)

Willbb234, others may know the techical details, but I have noted that articles remain in Category:Articles with long short description for some time after the SD has been shortened. Presumably it is only updated periodically. MichaelMaggs (talk) 22:55, 26 March 2021 (UTC)

MichaelMaggs when I shortened the SD at I SS Panzer Corps, it instantly removed the listing from the category. I likewise am not familiar with the technical details so we'll see what others have to say. Kind regards, Willbb234^Talk (please {{ping}} me in replies) 23:01, 26 March 2021 (UTC)

Willbb234 I'm fairly confident that it's due to shortdesc helper not recognizing the short description due to it containing a newline. Therefore it adds another short description instead of changing the old one. Since the old short description remains it will still categorize it as containing a long description. Ultimately only the active description should categorize it but that is technically infeasible (if not practically impossible without significant mediawiki changes). Galobtter do you want to take a look? --Trialpears (talk) 23:31, 26 March 2021 (UTC)

Yeah it's because of the regex not capturing new lines. Will fix. Galobtter (pingó mió) 20:28, 19 April 2021 (UTC)

Infobox			Transclusions	Suggested short description, based on WP:HOWTOSD	Status
Template:Chembox			11988	Chemical compound	Done
Template:Infobox award			10425	Award	Done
Template:Infobox book			46515	[year] book by [author]	Auto short description possible: not yet done
Template:Infobox character			7462	Fictional character by [author]	Auto short description possible: not yet done
Template:Infobox film			144829	[year] film by [director]	Auto short description possible: not yet done
Template:Infobox medical condition			8971	Medical condition	Done
Template:Infobox drug			77377	Drug	Not done Too many exceptions, where the infobox is used for things that most people wouldn't normally describe as a "drug"
Template:Infobox body of water			16676	Body of water in [country]	Auto short description possible: not yet done
Template:Infobox mountain			25657	Mountain in [country]	Auto short description possible: not yet done
Template:Infobox river			27682	River in [country]	Discussion started
Template:Infobox company			77502	Company in [country]	Auto short description possible: not yet done
Template:Infobox university			25758	University in [country]	Discussion started
Template:Infobox military conflict			18680	Military conflict [(dates)]	Not done. Probably best to add anything requiring date parsing either manually or by bot
Template:Infobox military unit			23693	Military unit	Discussion started
Template:Infobox organization			29468	Organization in [country]	Auto short description possible: not yet done
Template:Infobox airport			15480	Airport in [country]	Auto short description possible: not yet done
Template:Infobox road			23546	Road in [country]	Not done. Not easy, as conventions are too different in different regions
Template:Infobox station			50793	Station in [country]	Auto short description possible: not yet done
Template:Infobox football club			24683	Football club in [country]	Auto short description possible: not yet done

There are many infoboxes that still don't, but probably could, fill in missing short descriptions automatically. Here's a list of possibilities that have fairly heavy usage (though I don't know how many instances of each already have a manual description). There are other templates of course, but not all lend themselves to the creation of a simple SD. I'm happy for the table to be corrected/amended and used as a working document, if that would be helpful. MichaelMaggs (talk) 12:09, 29 March 2021 (UTC)

Thanks for doing this! I've had a brief look at the descriptions not depending on parameters since those are the easiest. I don't think "Drug" is a good description for {{Infobox drug}} since it's used on many thing basically no one would call drugs, such as Vitamin C and Dopamine. "Medical condition" seems fine for {{Infobox medical condition}}, but should check with WikiProject Medicine. {{Infobox military unit}} and "Military unit" may be fine but it would apply it to some paramilitary groups and such which isn't great, but those can likely be dealt with manually if that's the only issue. Award for {{Infobox award}} I can't come up with any major problems, but "Award" at something like 45th Annual Grammy Awards is just redundant. If anyone want help implementing these I'm happy to help. --Trialpears (talk) 09:45, 3 April 2021 (UTC)

Template:Infobox typeface seems like another low hanging fruit. "Typeface" or "[Category] typeface" seems very reasonable and uncontroversial. //Lollipoplollipoplollipop::talk 10:32, 18 April 2021 (UTC)

Lollipoplollipoplollipop, Yes that seems fairly straightforward for a template editor to do. It's perhaps not as high priority though, as others in the list as there are only 390 articles using that infobox that lack a short description. MichaelMaggs (talk) 12:08, 18 April 2021 (UTC)

I at least won't be taking on {{infobox road}}. Conventions are too different in different regions and the project would likely balloon in scope to accommodate different cases or only take care of a small minority of cases. My assessment is that it's probably possible to make it good, but it's not low hanging fruit. --Trialpears (talk) 20:03, 19 April 2021 (UTC)

I have found that "Fictional character from $series" is more appropriate in many cases (and more common right now). — Goszei (talk) 18:48, 17 May 2021 (UTC)

FYI

 – Pointer to relevant discussion elsewhere.

Please see: Template talk:Disambiguation page short description#Short?. — Goszei (talk) 04:40, 3 April 2021 (UTC)

For section "Auto-generated from infoboxes and other templates" wondering if "Infobox sports season" can be added as Done? Noticed within Category:Manitoba Junior Hockey League seasons that at 2007 and going forward, the auto-gen seems to be working. I'm new to this WP so not sure if explaining clearly. JoeNMLC (talk) 21:06, 12 April 2021 (UTC)

JoeNMLC,  Done editing tables is a bit fiddly. MichaelMaggs (talk) 15:52, 13 April 2021 (UTC)

The tracking category Category:Articles with long short description still seems to include articles that should not be there, which makes it hard to work through sytemematically. The very first article, 1 Train (song), is included even though the manual template {{Short description|2013 song by ASAP Rocky}} was added to it on 26 March, presumably to overrride the default description from {{Infobox song}}. Not sure what needs to be done to fix this. MichaelMaggs (talk) 12:27, 18 April 2021 (UTC)

The article has two descriptions — 2013 song by ASAP Rocky featuring Kendrick Lamar, Joey Badass, Yelawolf, Danny Brown, Action Bronson and Big K.R.I.T. from the infobox and 2013 song by ASAP Rocky from the local SD template. The infobox SD template adds the article to the category — GhostInTheMachine ^{talk to me} 13:02, 18 April 2021 (UTC)

Maybe the way to handle this would be to add |short_description= as a field in the infobox, which then prevents the infobox from even creating the default one. --Gonnym (talk) 13:04, 18 April 2021 (UTC)

See Wikipedia talk:WikiProject Short descriptions/Archive 3#Long SD tracking above on how this was fixed in {{infobox school}}. This should be added to all infoboxes that generate SDs. MB 15:53, 18 April 2021 (UTC)

Ah yes, thanks, that was it! Now all we need to do is to fix the infoboxes ... MichaelMaggs (talk) 20:15, 18 April 2021 (UTC)

If someone gives me a list of affected templates I could stop all infoboxes from generating descriptions longer than 100 characters, assuming no one objects here in the next few days. I don't think a short description parameter is a good idea though since that would mess with tools and harder for editors to know what is the preferred way to handle the situation. --Trialpears (talk) 20:42, 18 April 2021 (UTC)

It would be fantastic if you would edit the templates: they are all listed at Category:Templates that generate short descriptions. Maybe some some drop-in code or a nice module that makes it easy for template writers to do the right thing when creating new templates? Ideally there are two things to be done:

1. Using Template:Has short description, prevent the template from generating a short description at all if {{short description}} already exists in the wikicode

2. Adjust the logic to make sure that any short description generated is never greater than the recommended 40 characters.

There shouldn't be a need to make any changes specific to 100 characters, as 1. and 2. will deal with the issue. 1. in particular should prevent all sorts of future confusion. MichaelMaggs (talk) 06:45, 19 April 2021 (UTC)

That's a good idea! I made a template named {{Template short description}} which checks {{Has short description}} and the length like you suggested, but also has support for Module:Is infobox in lead if a pattern is provide. --Trialpears (talk) 18:17, 19 April 2021 (UTC)

Trialpears, that looks excellent. My only concern is the provision of a user-selectable cutoff parameter, which editors are likely to set at much too high a value, without any regard to the 40 character max recommendation. Bearing in mind that an infobox coded to create a short description is producing a default value, which is likely to be extensively seen, used, and copied, it really ought to comply with the guidance at WP:SDCONTENT. I'd prefer no adjustable parameter, with 40 chars max hard-coded. A user finding that their overly-long text doesn't create the expected description is then incentivised to shorten it, or to add a description with the manual template. MichaelMaggs (talk) 14:24, 20 April 2021 (UTC)

MichaelMaggs, I would trust template editors not to put something crazy large in there and don't think restricting it is necessary, but 100 is a very high default value. On the other hand I'm uncertain if putting the limit at 40 would actually be beneficial since that would remove a lot of serviceable, but suboptimal, descriptions. It's ultimately a trade off and there isn't a right answer, but my feeling after some thinking is that somewhere 50-60 probably is quite good. --Trialpears (talk) 14:50, 20 April 2021 (UTC)

Trialpears, any default of more than 40 would concern me. It's pretty clear that most editors have no idea there's a consensus of no more than about 40, nor the reason for that: cut-off problems that become increasingly severe above that point (and indeed below, but 40 was a reasonable consensus figure). Making a new template that defaults to an undiscussed value of 50-60 effectively overrides the consensus figure and encourages template editors to make descriptions which are too long/complicated for mobile views. What about a default of 40, a reference to WP:SDFORMAT in the documentation to discourage higher values being set, but allowing template editors to go higher if they really have to? MichaelMaggs (talk) 11:42, 22 April 2021 (UTC)

MichaelMaggs, makes sense. If we see massive drops in generated descriptions we have to investigate further. I've set the default to 40 chars and updated the documentation. --Trialpears (talk) 19:27, 22 April 2021 (UTC)

Trialpears, looks good to me. MichaelMaggs (talk) 19:44, 22 April 2021 (UTC)

Good news! PearBOT 5 recently finished its 100 thousand+ edit long run of adding short descriptions to biographies based on the first sentence in the lead. There is however more potential for this generation method and by loosening on some restrictions in the first run, such as going from always requiring the first word of the description being a nationality to only requiring that a nationality is present and allowing it to add descriptions to politicians articles which wasn't done before due to it not being able to with 100% confidence figure out their term. Now it gives them a description without any dates. There are also a few more possible cut off points which makes more descriptions fall under the 40 character limit and generates more descriptions. Here we have 300ish descriptions generated by the new bot. The plan is to start this run next weekend if no issues arise. There is also a thread at Wikipedia:Bots/Noticeboard#Second run. --Trialpears (talk) 20:09, 25 April 2021 (UTC)

Albert I of Belgium is in Category:Short description matches Wikidata and in Category:Short description is different from Wikidata. In fact, according to Petscan, we have no less than 33,706 articles with this situation. Fram (talk) 11:07, 28 April 2021 (UTC)

Generally this is because articles have two or more short descriptions. Albert has two — Third King of the Belgians from the local template and King of the Belgians from the infobox. The local one is the "live" one and it matches Wikidata — GhostInTheMachine ^{talk to me} 11:45, 28 April 2021 (UTC)

It is not an error to have multiple short descriptions, that is what the {{{noreplace}}} is for — GhostInTheMachine ^{talk to me} 15:48, 28 April 2021 (UTC)

Fram, I think this is essentially the same error that I raised in the section above, #Error in the tracking category Articles with long short description. The solution is for somebody to amend {{Infobox royalty}} to prevent it generating a default description if {{short description}} is present elsewhere on the page. Trialpears has a template {{Template short description}} which will do that. See also the brief notes here. MichaelMaggs (talk) 13:03, 28 April 2021 (UTC)

I feel like we've waited long enough without objections to deploy {{Template short description}}. I don't have time to take care of it today, but tomorrow I plan on doing a wide rollout including {{Infobox royalty}}. Then we can have a look at the Petscan and find out how many are left. --Trialpears (talk) 13:24, 28 April 2021 (UTC)

What is implied here by deploy? — GhostInTheMachine ^{talk to me} 15:45, 28 April 2021 (UTC)

That I plan to edit several templates to use {{Template short description}} instead of {{short description}} to prevent issues like this and generation of longer than recommended descriptions. This is discussed at #Error in the tracking category Articles with long short description, but if you have any concerns I will of course not do so without resolving it. --Trialpears (talk) 20:08, 28 April 2021 (UTC)

If the purpose is to reduce the chance of an infobox creating a short description that is too long, then we should probably explore creating a template named something like {{Infobox short descripion}} that is designed to only function within an infobox and takes a list of several alternative candidate short descriptions generated by the infobox code. The infobox can then start with the original full size text and fail gracefully to progressively simpler versions. The template then emits the first of these that is "short enough", perhaps with the final option being no text at all. BTW – Starting a template name with template seems slightly evil to me. — GhostInTheMachine ^{talk to me} 20:53, 28 April 2021 (UTC)

Are there any sandboxes or tests for this? — GhostInTheMachine ^{talk to me} 20:53, 28 April 2021 (UTC)

There are two main purposes. To avoid having double descriptions by checking if one is defined explicitley solving categorization issues like having pages in both Category:Short description matches Wikidata and in Category:Short description is different from Wikidata or a page being categorized as having a long short description when it doesn't and to avoid longer than recommended descriptions to be generated. Having a more clever template that can try to deal with too long descriptions is of course preferable and is done in places, such as {{Infobox school}}, but many times it is not feasible. It was not too well tested, but I've now set up testcases as well as preview tested it with more realistic {{Infobox river/sandbox}}. Before every edit to an actual used template there would be more tests as well. The name is like this because there are other templates than infoboxes generating descriptions. It doesn't sound great though. I've considered "automatic short description" but that makes it sound like something more sophisticated and something like "Check short description" which doesn't sound great either, but I guess it works. --Trialpears (talk) 22:51, 28 April 2021 (UTC)

Wikipedia desktop searches are now being enhanced with a short description under each title in the drop-down list. Did that happen very recently? I could have sworn those short descriptions weren't there a few days ago. MichaelMaggs (talk) 20:56, 28 April 2021 (UTC)

Which skin do you have set? — GhostInTheMachine ^{talk to me} 21:12, 28 April 2021 (UTC)

GhostInTheMachine, Vector (the latest version). Selecting the "Use Legacy Vector" option makes the short descriptions go away. I may be unusual in quite liking the new Vector skin, though it does odd rather things on my wide screen monitor - WONTFIX unfortunately. MichaelMaggs (talk) 21:21, 28 April 2021 (UTC)

It looks like a new feature of mw:Reading/Web/Desktop_Improvements/Features/Search was unleashed on the English Wikipedia. MichaelMaggs (talk) 21:29, 28 April 2021 (UTC)

Sadly, all the new toys are going into Evil Vector — GhostInTheMachine ^{talk to me} 21:33, 28 April 2021 (UTC)

I've just stumbled across for the first time a See also section that makes use of {{annotated link}}: see Christmas#See_also. It looks very good. MichaelMaggs (talk) 15:24, 15 May 2021 (UTC)

That looks very nice; saves a lot of time! Sdrqaz (talk) 17:50, 15 May 2021 (UTC)

I propose that Wikipedia talk:Short description be merged to here. These pages are both for discussion of short descriptions and should be consolidated. {{u|Sdkb}} ^talk 19:56, 23 May 2021 (UTC)

• Support as proposer. {{u|Sdkb}} ^talk 19:56, 23 May 2021 (UTC)
• Support. There's no practical distinction between the two talk pages. MichaelMaggs (talk) 20:36, 23 May 2021 (UTC)

As suggested below, merging to the other page would be better. MichaelMaggs (talk) 09:27, 24 May 2021 (UTC)

• Support. Completely reasonable; was wondering when it would happen. Sdrqaz (talk) 22:10, 23 May 2021 (UTC)

  For what it's worth, I'm fine with it being merged the other way. Sdrqaz (talk) 10:29, 24 May 2021 (UTC)

• Support a merger, but maybe in the other direction? Short descriptions will be here long term. But the project could become much less active at some point. And note that Template talk:Short description already redirects there. MB 23:18, 23 May 2021 (UTC)
• Do not merge to here. The standard should be that interested people follow the central page which is Wikipedia:Short description. A typical wikiproject is Wikipedia:WikiProject Music which is for people interested in a maintaining articles related to music. Wikipedia talk:Short description has 139 page watchers, 69 of whom visited recent edits. For Wikipedia talk:WikiProject Short descriptions, the numbers are 100 and 50. Johnuniq (talk) 23:34, 23 May 2021 (UTC)
• Merge in the other direction. – Jonesey95 (talk) 23:48, 23 May 2021 (UTC)
• Support as someone who's just been confused by there being two talk pages. - X201 (talk) 07:35, 24 May 2021 (UTC)
• Support a merge as, like X201, I hadn't realized there were two pages, and could never find discussions I knew I had seen. But merge in opposite direction, as that seems the appropriate place for SD topics. Project discussions must (with time) be the smaller number, surely. — JohnFromPinckney (talk) 15:22, 24 May 2021 (UTC)
• Merge in the other direction. I never knew this page was here. The other page has more archives, also. --Whiteguru (talk) 07:34, 9 June 2021 (UTC)
• Merge, preferably in the other direction, per MB and Whiteguru. 156.57.13.133 (talk) 00:40, 9 August 2021 (UTC)

There seems to be consensus to merge this talk page into Wikipedia talk:Short description. I'd be happy to do it, but would like to check the procedure. Would it work to (1) archive all discussions on this page, then (2) turn this page into a redirect? What should happen to the archives? Should they be left where they are, with a note on Wikipedia talk:Short description pointing out where they can be found? MichaelMaggs (talk) 15:47, 20 September 2021 (UTC)

@MichaelMaggs: This all makes sense. Regarding archives, I think you can just mirror what as done about Template talk:Short description, whose archives were just copied as a "dead" archive box at WT:SHORTDESC. JBchrch talk 18:08, 20 September 2021 (UTC)

OK, preparing to redirect this page now to Wikipedia talk:Short description. MichaelMaggs (talk) 12:29, 24 September 2021 (UTC)

When using the "none" option on an article, the "Missing page description" warning still shows at the top of the article. Shouldn't this be suppressed to avoid confusion? - X201 (talk) 07:32, 24 May 2021 (UTC)

Definitely. Just to be clear this is the "Show page description" gadget. It doesn't recognize that there delibrettly isn't a description since the actual description field is unchanged. (Really it would be optimal with a software change to deal with that) The page that would have to be edited is MediaWiki:Gadget-Page descriptions.js, I'm not particlarly good at writing scripts and if someone else could deal with it that would be great. --Trialpears (talk) 08:42, 24 May 2021 (UTC)

The article Sodium sulfate has a short description of Chemical compound with formula Na<sub>2</sub>SO<sub>4</sub>. The view of an editor was that simple formatting HTML tags would be acceptable as they display correctly.
Currently, WP:SDFORMAT says Each short description should ... be written in plain text – without HTML tags or Wiki markup. I assume it is correct that short descriptions should not contain wiki markup (especially links and templates), but what about simple HTML tags?

1. For mobile and in all apps, are simple HTML tags harmlessly removed or do they just display correctly?
2. If short descriptions may contain a limited set of simple HTML tags, how do we identify and define simple? — GhostInTheMachine ^{talk to me} 18:14, 22 June 2021 (UTC)

It's pretty clear to me that written in plain text – without HTML tags or Wiki markup means that "simple" HTML tags are not acceptable. JBchrch talk 18:40, 22 June 2021 (UTC)

I looked at that SD in two different views, and the sub tags are stripped before the SD is displayed. They appear to be stripped by the rendering engine, so they should probably just be removed. I think that our guidance is valid. – Jonesey95 (talk) 19:16, 22 June 2021 (UTC)

Why not replace them with the actual subscript characters if the concern is having visible HTML tags? E.g.: Na₂SO₄ -> Na₂SO₄ //Lollipoplollipoplollipop::talk 02:34, 23 June 2021 (UTC)

Why does Netherlands in the Junior Eurovision Song Contest 2015 have "Musical artist" as the short description? Thanks. Mike Peel (talk) 18:56, 4 July 2021 (UTC)

Because it contains {{infobox musical artist}}, perhaps a questionable choice for that article. It can be overridden with a local SD template in the article. – Jonesey95 (talk) 19:02, 4 July 2021 (UTC)

There is edit-warring over description, which is nearly identical to the article's title. More at Wikipedia talk:Short description#Description repeating article's title. --Triggerhippie4 (talk) 21:30, 28 July 2021 (UTC)

Reached as of this timestamp with 3,620,088 pages. Sammi Brie (she/her • t • c) 04:23, 22 August 2021 (UTC)

On this page I am repeatedly running into an issue where a short description is entered, and the page reports Wikimedia template. Any effort to correct this fails. How do we sort this and get the correct short desc into the page? (should be "Asian rice dish") --Whiteguru (talk) 05:43, 7 September 2021 (UTC)

Also potentially noteworthy is that the page did not have any local short description until July 25 when Whiteguru added "Traditional dinner dish...". Yet just one day later on July 26, Pi bot over at Wikidata made an edit saying "Importing description from enwiki: wikimedia template". But in between those two times, the short description of claypot rice given in the template did not change; the only edit made was a minor copyedit by myself. ChromeGames923 (talk · contribs) 05:57, 7 September 2021 (UTC)

Ping to bot-op Mike Peel. --Trialpears (talk) 09:00, 7 September 2021 (UTC)

Note that the bot runs on Mondays and Fridays, so the day the bot edits may not be an exact indication of when the short description was added. I've added 'wikimedia template' to the list of excluded short descriptions, but it sounds like the problem has been fixed below anyway. Thanks. Mike Peel (talk) 10:10, 7 September 2021 (UTC)

Fixed, {{Singapore-cuisine-stub}} had a badly placed short description that appeared on articles and couldn't be overridden. --Trialpears (talk) 07:19, 7 September 2021 (UTC)

Thanks for fixing it, that's a good catch! On that note, do you know if shortdesc helper safely deals with adding short descriptions on template pages? ChromeGames923 (talk · contribs) 08:53, 7 September 2021 (UTC)

I believe it's just disabled in template space. It's quite unnecessary adding descriptions to templates in my opinion. --Trialpears (talk) 08:58, 7 September 2021 (UTC)

Thanks, Trialpears. A good explanation. --Whiteguru (talk) 10:05, 7 September 2021 (UTC)

I made a similar fix to {{Cambodia-cuisine-stub}}. Whilst seeking other cases, I found other SDs in templates without noinclude or noreplace which might override explicit SDs at the top of articles. Are they cause for concern? (The rough search has false positives and negatives but still seems useful.) Certes (talk) 10:29, 7 September 2021 (UTC)

A lot of these should indeed be noreplace'd or noincluded. Beware of exceptions such as doc pages though. --Trialpears (talk) 12:16, 7 September 2021 (UTC)

Many are OK. Any template setting the SD should also have a {{Auto short description}} template in the doc page – which is a much trickier search — GhostInTheMachine ^{talk to me}

Hi, I've been writing an automatic short description for this template (see the talk page). Could someone experienced in writing automatic short descriptions check to see if I've made any mistakes? Thanks! ― Qwerfjkltalk 19:59, 19 September 2021 (UTC)

I noticed straight away from your example at User:Qwerfjkl/sandbox3 that you don't fully comply with WP:SDDATES, specifically "Care should be taken where WP:BLP applies, and birthdates for living people should not be copied from the page unless sourced." As sourcing is hard to check, it's safest to avoid birth dates for living people entirely. MichaelMaggs (talk) 21:12, 19 September 2021 (UTC)

 Done ― Qwerfjkltalk 06:16, 20 September 2021 (UTC)