Did the bot screw something up? Suggestions? Leave a message on this page and I'll get back to you. Headbomb {talk / contribs / physics / books} 20:13, 5 September 2011 (UTC) Reply

Archives

Archive 1

This page has archives. Topics inactive for 180 days are automatically archived 1 or more at a time by Lowercase sigmabot III if there are more than 4.

title=none works only on journal citations. If the bot is trying to change one of those to a book citation, as it did here, something is wrong. (In this case, what is wrong is that some other broken bot previously added bogus doi and isbn tags to the citation.) Here is another bad edit of the same flavor: it really is a journal cite, to a book review, on which another bad bot has tacked on the book isbn. —David Eppstein (talk) 18:02, 22 June 2026 (UTC)Reply

Well, not much I can do about GIGO. Headbomb {t · c · p · b} 18:04, 22 June 2026 (UTC)Reply

Excuses, excuses. A bot borked the citation and then another bot borked it more. It was not borked before the bots started on it. Anyway, yes, there is something you can do, at least on the first one: flag a problem when you see title=none rather than continuing to break things. —David Eppstein (talk) 18:07, 22 June 2026 (UTC)Reply

Another bad edit that I can't figure out where the isbn came from but it really is a journal cite: [1]. —David Eppstein (talk) 18:07, 22 June 2026 (UTC)Reply

Yet another with title=none. So far roughly half of the bot edits that I've checked have been worsenings rather than improvements. That is an unacceptable error rate. —David Eppstein (talk) 18:15, 22 June 2026 (UTC)Reply

Another with title=none and another without. Rate of bad bot edits creeping up, maybe a majority? —David Eppstein (talk) 18:17, 22 June 2026 (UTC)Reply

And another one that really is a journal issue with an isbn (look at the Google Books preview page metadata). —David Eppstein (talk) 18:19, 22 June 2026 (UTC)Reply

BTW some edit got lost here in an edit conflict I think, I'm skipping the

cases and will do those manually later. Headbomb {t · c · p · b} 22:59, 22 June 2026 (UTC)Reply

In this edit the bot changed a double-bracketed url (rendering as a link in brackets) to a single-bracketed url (rendering as the link without the brackets around it). Why is the bot making sort-of-cosmetic edits like this at all? It was adjacent to but not actually in a citation template. —David Eppstein (talk) 18:25, 22 June 2026 (UTC)Reply

Part of AWB genfixes. Headbomb {t · c · p · b} 18:26, 22 June 2026 (UTC)Reply

It was not a fix. It was a change to appearance. Why is this bot even running AWB? What does it have to do with its charge to clean citations? —David Eppstein (talk) 18:27, 22 June 2026 (UTC)Reply

gigo; no pipe characters in bracketed external links:

[[https://archive.org/details/conceptsofmodern0000stew|Alt URL]] → [URL]

gives this broken url: https://archive.org/details/conceptsofmodern0000stew%7CAlt

replacing the pipe with a space character gives:

[[https://archive.org/details/conceptsofmodern0000stew Alt URL]] → [Alt URL]

and give this working url: https://archive.org/details/conceptsofmodern0000stew

—Trappist the monk (talk) 18:37, 22 June 2026 (UTC)Reply

Stop using GIGO as a bad excuse for bad bot behavior. All of the articles the bot touches can be claimed to be GI on the basis that they already have a journal citation with an isbn. Nevertheless the bot should not be making them worse. —David Eppstein (talk) 22:14, 22 June 2026 (UTC)Reply

It's exposing already-existing problems, it's not making articles worse. Headbomb {t · c · p · b} 22:51, 22 June 2026 (UTC)Reply

Some changes are mildly though pretty trivially helpful. Some changes are more or less neutral, taking one incorrect version and replacing it by a slightly different incorrect version. Some changes make the citation templates more wrong than they started and make the articles worse. From my skimming, these three categories seem to be about in equal proportions. That's nowhere close to acceptable for a large scale bot action. The result of this is not to "expose already-existing problems", unless you hire a human to manually double check every bot edit, which as far as I can tell you are not doing. –jacobolus (t) 00:48, 23 June 2026 (UTC)Reply

Not an exaggeration: All of my editing time today, with the exception of time set aside for creating a new article, was spent running around after the bot fixing messes that it created. One more similarly wasted day of human editing time and I will block the bot. This is far from ready for prime time. The decision of which direction to fix these problem references is nontrivial and the bot is far too often making things worse rather than better. Stop running it. —David Eppstein (talk) 04:26, 23 June 2026 (UTC)Reply

"messes that it created". The bot created no mess. To my knowledge, every "fix" you made after this bot fixed an issue that was already present in the article, just not evident. Headbomb {t · c · p · b} 06:28, 23 June 2026 (UTC)Reply

The bot has been making maybe 1/3 of the citations it touches better and maybe 2/3 worse. It has taken citations that had a flag for a problem, and turned them into citations with worse problems and no flag. That is not helpful behavior. If a human editor were behaving this badly they would be blocked already. —David Eppstein (talk) 07:14, 23 June 2026 (UTC)Reply

By my count, the type of edits you complain about are on the order of one per 400. Given that these are already borked, I don't see how the bot is making things worse. Headbomb {t · c · p · b} 07:15, 23 June 2026 (UTC)Reply

I'm counting what I see on my watchlist and it is far far far above 1 in 400. More like 2 in 3. And yes, making citation metadata even less accurate while simultaneously masking the issue that triggered a warning flag is making things worse. But also, making bad edits that waste human editor time to the point that they must either let bad edits stand or have no time to do anything else is making things worse. —David Eppstein (talk) 07:20, 23 June 2026 (UTC)Reply

And it put the book title under "chapter. special:diff/1360658728 to Fibonacci sequence. –jacobolus (t) 20:53, 22 June 2026 (UTC)Reply

Lol. In special:diff/1360653250 to Eigenvalues and eigenvectors the bot put: "title=New York: McGraw-Hill". –jacobolus (t) 21:15, 22 June 2026 (UTC)Reply

In special:diff/1360613972 to Astrolabe the bot again changed a book title to 'chapter' and listed the book series as 'title'. –jacobolus (t) 21:32, 22 June 2026 (UTC)Reply

It's very weird for the "Citation Cleaner Bot" to be changing template shortcut names, e.g. from {{also}} to {{see also}} or from {{slink}} to {{section link}}. In my opinion these edits are unhelpful, and as far as I can tell there is no editor consensus to perform them. In particular, the name "slink" is explicitly listed as a shortcut name at the top of Template:Section link/doc; my impression is that the general consensus is that such shortcut names are considered acceptable long-term alternate names, and should not be changed by bots. But in any event, such changes should not be mixed around with citation changes. –jacobolus (t) 20:55, 22 June 2026 (UTC)Reply

Frankly I'd urge the bot to be stopped altogether until this is sorted. These kinds of edits are not mentioned at all in Wikipedia:Bots/Requests for approval/CitationCleanerBot 7. –jacobolus (t) 21:00, 22 June 2026 (UTC)Reply

The bot was trialled and approved with AWB genfixes. I see no reason to not make AWB genfixes as part of the bots task. Headbomb {t · c · p · b} 22:53, 22 June 2026 (UTC)Reply

Many of these changes are not "fixes". Can you explain where there was a request/approval for a bot to rename templates, move wiktionary link boxes around, etc.? I'm not seeing it. I don't think there's any widespread Wikipedian demand for such a bot. If someone did approve this, in my opinion they screwed up, and the bot approval should probably be rescinded altogether and this person should do a better job checking next time.

To be explicit, the page says:

Function details: Basically, if a {{citation}} has as |work= and |isbn= set, the bot performs these parameter swaps/renamings
|title= -> |chapter=
|work= - > |title=
|url= -> |chapter-url=
|url-access= -> |chapter-url-access=
Would only affects articles in Category:CS1 maint: work parameter with ISBN.

That description seems to be completely inaccurate. Ping @user:Primefac. –jacobolus (t) 22:58, 22 June 2026 (UTC)Reply

I'm happy to update the BRFA to indicate that genfixes are enabled if that makes things better. Primefac (talk) 01:41, 23 June 2026 (UTC)Reply

What's the justification for mixing these entirely unrelated cosmetic edits into a "Citation cleaning" task? –jacobolus (t) 01:57, 23 June 2026 (UTC)Reply

Because they're general fixes for a page? From a BAG perspective, unless an edit is going to be lost in the genfixes (e.g. this task which fixed some double-spacing issues) then someone running an AWB bot might as well do the genfixes along with the main task, because someone will eventually hit the page with genfixes enabled anyway. Primefac (talk) 10:51, 23 June 2026 (UTC)Reply

These are not "fixes". They arbitrary changes unsupported by editor consensus. –jacobolus (t) 15:29, 23 June 2026 (UTC)Reply

Anyway though, the only options for this bot are either (a) a human editor carefully "review[s] every edit" to make sure that it follows editor consensus and doesn't make mistakes as demanded by the AWB rules of use, or (b) the the bot does something narrowly constrained to a clear task which is nearly impossible to mess up, so that human editors don't have to spend inordinate effort cleaning up after it. "Someone is going to make a semi-automated mess so I might as well just do it myself" is really shoddy reasoning and this approach evades responsibility and abuses the rules. –jacobolus (t) 15:36, 23 June 2026 (UTC)Reply

If there are issues with the genfixes themselves, bring it up at WT:AWB. End of story. Primefac (talk) 23:11, 23 June 2026 (UTC)Reply

I'm not stopping anyone from discussing at whatever AWB page. But regardless, You folks need to follow Wikipedia bot policy and the AWB "rules of use", which you are currently refusing to do. –jacobolus (t) 23:46, 23 June 2026 (UTC)Reply

Why did the bot remove the Wiktionary link from the Etymology section in special:diff/1360624360 to Calculus? That's the most relevant and reader-useful place for it. Why is a "Citation cleaner" bot deciding where to put Wiktionary links? –jacobolus (t) 21:11, 22 June 2026 (UTC)Reply

I am spending all my editing time today running around after the bot cleaning up after its bad edits on my watchlist (seemingly over half of the edits I check). Some of the edits are valid conversion to a book citation. Some are conversions that should be valid but that it gets wrong (usually mistaking a series title for a book title). Some are journal articles in journal issues that happen to have isbns. Some are the bot making worse citations that other bots have already broken (often, references to book reviews to which another bot has mistakenly added book metadata such as an isbn). Some are I don't know what but wrong. GIGO is no excuse because the GI part of this is automatically present in what the bot is supposed to be cleaning up. With this low a hit rate the bot should not be operating. These cleanups are apparently too complex for a bot to decide correctly among options for how to address them. The bot should be stopped. —David Eppstein (talk) 22:12, 22 June 2026 (UTC)Reply

If you have some specific examples of where the bot is not performing the task (specifically, of updating the CS1 stuff) properly, please provide them to that any procedural or programmatic errors can be corrected. Primefac (talk) 01:41, 23 June 2026 (UTC)Reply

Part of AWB genfixes. If this bot didn't do this, another would. Headbomb {t · c · p · b} 23:03, 22 June 2026 (UTC)Reply

If this is considered a "fix" then AWB is broken and should not be used. –jacobolus (t) 23:06, 22 June 2026 (UTC)Reply

Per WP:AWBRULES:

You are responsible for every edit made. You are expected to review every edit, just as if you were making an edit using Wikipedia's edit form when editing by hand. Do not sacrifice quality for speed, and review all changes before saving.
Do not make controversial edits with it. Seek consensus for changes that could be controversial at the appropriate venue; village pump, WikiProject, etc. "Being bold" is not a justification for mass editing lacking demonstrable consensus. If challenged, the onus is on the AWB operator to demonstrate or achieve consensus for changes they wish to make on a large scale.
Do not make insignificant or inconsequential edits. An edit that has no noticeable effect on the rendered page is generally considered an insignificant edit. If in doubt, or if other editors object to edits on the basis of this rule, seek consensus at an appropriate venue before making further similar edits.

–jacobolus (t) 23:09, 22 June 2026 (UTC)Reply

I see neither controversial, nor insignificant nor inconsequential edits made. Headbomb {t · c · p · b} 23:33, 22 June 2026 (UTC)Reply

Can you explain what the justification is for moving the wiktionary link in Calculus? This seems clearly controversial to me.

The template documentation says:

The template may be placed anywhere, such as the External links section, the beginning of the article, or in the article's etymology section if one exists. It may also be placed inline.

Nowhere does it say anything remotely similar to "all use of this template outside the external links section is banned" or "bots should move this template to the external links section". –jacobolus (t) 00:10, 23 June 2026 (UTC)Reply

Again, AWB genfixes. If this is wrong, ping AWB maintainers. Headbomb {t · c · p · b} 00:13, 23 June 2026 (UTC)Reply

The bot operator is responsible for all edits. If they are controversial they must be paused until the script/tool is fixed or until you have demonstrated consensus for the changes. To quote again, "If challenged, the onus is on the AWB operator to demonstrate or achieve consensus for changes they wish to make on a large scale." –jacobolus (t) 00:17, 23 June 2026 (UTC)Reply

Can you explain what the justification is for changing the template name "{{slink}}" to "{{section link}}" everywhere, despite "slink" being explicitly listed in the template documentation as a blessed shortcut? There is no policy or Wikipedian consensus that I can see for bots making mass changes to recognized template shortcut names, and this seems both (a) inconsequential in the sense that it does not affect output, and (b) harmful in that it bloats article markup and is pointlessly distracting. I say it is a controversial change, which should not be snuck in as part of "Citation cleaning". –jacobolus (t) 00:16, 23 June 2026 (UTC)Reply

Again, AWB genfixes. If this is wrong, ping AWB maintainers. Headbomb {t · c · p · b} 00:22, 23 June 2026 (UTC)Reply

Again, you have responsibility for this. If there is consensus for such changes, the onus is on you to demonstrate it. If you do not have consensus for such changes, then you must immediately stop making more of them until you have achieved consensus or fixed your broken tool. See also Wikipedia:Fait accompli. You are making a tremendous mess for other people to deal with, and then disclaiming any personal responsibility. –jacobolus (t) 00:24, 23 June 2026 (UTC)Reply

I'm going to be somewhat pedantic here, but there is consensus for AWB genfixes to be making these edits purely by the dint of then being part of the genfixes. Thus, as Headbomb says, if you have an issue with the genfixes, you need to get those genfixes altered. Primefac (talk) 01:36, 23 June 2026 (UTC)Reply

That being said (and I apologise for making a dozen edits here, it's rather "late" for me but I'm travelling and trying to keep myself awake) the issue with {{slink}} is actually something that can be edited by anyone simply by removing it from WP:AWB/TR; I can definitely see the argument for leaving that be, similar to how {{ill}} is not on the list for renaming to {{interlanguage link}}. Primefac (talk) 01:45, 23 June 2026 (UTC)Reply

I did make that change in 2023, and was repeatedly reverted. There's a small group of editors who believe themselves to own that AWB page who will edit war to stop any changes they personally disagree with, and they don't care at all about the consensus among the wider project. Making any progress with them would be a large and obnoxious political fight. –jacobolus (t) 01:52, 23 June 2026 (UTC)Reply

I don't really have a good solution for you if you've tried already, other than bringing it up for broader discussion. Primefac (talk) 10:41, 23 June 2026 (UTC)Reply

Where is the demonstrated consensus for all of these various changes? Please give a link where there was a discussion about moving Wiktionary boxes. –jacobolus (t) 01:51, 23 June 2026 (UTC)Reply

I highly doubt there is one thread that lists all of these various issues, but for example moving sister project boxes is part of the MOS; that being said "should" doesn't mean "must" so there are clearly exceptions, but when AWB's genfixes were being designed I suspect the MOS was one of the places they looked to get their fixes. If there are specific concerns with specific changes, bring them up at the AWB talk; while I do note the big thread at AWB/TR's talk page is massive (and we've all participated in it) they don't all end that way. Primefac (talk) 10:41, 23 June 2026 (UTC)Reply

The {{wiktionary}} box documentation explicitly recommend putting it "in the article's etymology section". If that is now disallowed by the manual of style, it should probably be changed. Was there a discussion about it? –jacobolus (t) 15:31, 23 June 2026 (UTC)Reply

That is a genfixes issue and unrelated to this bot task. A conflict between instructions means that a discussion needs to be had to determine which instructions need to be updated. As mentioned, though, the MOS is not a mandate but it does inform things like AWB genfixes. If you think an explicit carve-out should be made for the wiktionary template, start a discussion at WT:AWB. If you think the MOS needs to be changed, bring it up on the MOS talk. Primefac (talk) 23:11, 23 June 2026 (UTC)Reply

I don't think this line in the MOS supports having a bot automatically move every instance of this template. Such an automatic action needs demonstrated editor consensus. If there is such a consensus, point to the discussion. (If there is a consensus for such an automatic move, then it should be done by a different bot with clear and separate approval, so that people can have a discussion about it someplace topical, instead of at the page of a bot ostensibly responsible for "citation cleaning".) If there is no such consensus, then please desist from doing such changes as part of any automatic process.

Lumping in totally unrelated bot changes like this is socially hostile – for example, it requires editors auditing this bot's activity to check on more than one type of thing, and it makes the edit summaries for the bot edits grossly inaccurate – and from what I can tell is unsupported by policy. –jacobolus (t) 23:47, 23 June 2026 (UTC)Reply

Is there a reason I am seeing the same edits from both CitationCleanerBot and User:JCW-CleanerBot? Shouldn't a constrained bot task like this be consolidated under a single bot name? —David Eppstein (talk) 23:18, 22 June 2026 (UTC)Reply

Absolutely my bad. These should have been performed under this account. I had a password issue earlier and I must have logged in the wrong account by accident. I'll be reviewing those. Headbomb {t · c · p · b} 23:37, 22 June 2026 (UTC)Reply

After coming back to my watchlist this morning only to see another five mistakes (out of 13 citations, an error rate still well above 1/3) I have blocked the bot. This error rate is far too high to be acceptable. For reference, the bad edits I found were:

• Tight span: wrong choice of title=
• Triangulation (topology): two bad citations; both were books in a series not a chapter in a book
• Wagner's theorem: wrong choice of title=
• Θ10: article in festschrift journal issue with isbn

—David Eppstein (talk) 18:17, 23 June 2026 (UTC)Reply

Actually the mistake rate is worse than that because I neglected to count another one on my watchlist already undone by another editor, sexagesimal (another case where journal= should have been series=). —David Eppstein (talk) 18:22, 23 June 2026 (UTC)Reply

Always exciting: a butchered citation from four years ago was extra butchered today. --JBL (talk) 00:14, 24 June 2026 (UTC)Reply

I've said it before, but: long ago I saw an art installation in which someone had modified a Xerox copier to make small modifications to the copied scan, randomly mixing in parts of another image, so that, as you made repeated copies of copies, all documents would eventually converge to a picture of Zippy the Pinhead. That's what this repeated polishing of citations and concomitant amplification of old mistakes most reminds me of. —David Eppstein (talk) 00:46, 24 June 2026 (UTC)Reply

Headbomb, given this task's apparent error rate, have you considered trying to run it as a supervised task with AWB's general fixes turned off? You would be able to see when it was proposing invalid edits of the kinds reported above, or GIGO edits that would need follow-up. – Jonesey95 (talk) 04:57, 25 June 2026 (UTC)Reply

Given edits [2] and [3] by Headbomb correctly fixing articles on my watchlist that this task would have incorrectly fixed, I was assuming that's what they were already doing. —David Eppstein (talk) 05:08, 25 June 2026 (UTC)Reply

It's a rather small error rate, which apparently disproportionally affects articles on David's watchlist, and I've already been skipping most of what could be predicted to go wrong, or corner cases I knew the bot couldn't handle. What I can't skip or predict is when something has been borked by something else. The block is rather unneeded at this point anyway, since the rest will be handled semi-automatically. Headbomb {t · c · p · b} 05:11, 25 June 2026 (UTC)Reply

I think it far likelier that it has a consistently high error rate and that the part that is an outlier is that I actually vet the bot edits on my watchlist rather than ignoring them and assuming they'll be fine. —David Eppstein (talk) 17:20, 25 June 2026 (UTC)Reply

You did what, about 30 reverts? Let's assume for sake of discussion that they were bad edits, across 22000 articles, that's an error rate of 1 in 733. Headbomb {t · c · p · b} 18:16, 25 June 2026 (UTC)Reply

It was across a far smaller set of articles, the ones that appeared on my watchlist. I didn't try checking other contributions of the bot. I think it far likelier that it has a consistently high error rate and that the part that is an outlier is that I actually vet the bot edits on my watchlist rather than ignoring them and assuming they'll be fine. —David Eppstein (talk) 01:04, 26 June 2026 (UTC)Reply

How about we make it go on demand, not whenever? _{In solidarity} Wikipedian12512 (Talking is fine | contribs) 00:06, 26 June 2026 (UTC)Reply

Agreed, this could be modified to use citation bot-like functionality as well as a preview or diff of the edit to be conducted. 🪐Kepler-1229b | talk | contribs🪐 01:21, 1 July 2026 (UTC)Reply

It can't, it's an AWB bot, not a web applet running on toolforge. Headbomb {t · c · p · b} 02:01, 1 July 2026 (UTC)Reply

nevermind then. 🪐Kepler-1229b | talk | contribs🪐 02:35, 1 July 2026 (UTC)Reply

"the part that is an outlier is that I actually vet the bot edits on my watchlist rather than ignoring them and assuming they'll be fine" confirm that I didn't check any of the edits on my watchlist, so there is definitively the potential. Jo-Jo Eumerus (talk) 12:33, 1 July 2026 (UTC)Reply

@David Eppstein: mind unblocking this? Pretty sure I cleaned up everything borked citation manually. The issue seemed to have been stemming from MR reviews, which I can easily skip. Headbomb {t · c · p · b} 19:28, 24 July 2026 (UTC)Reply

Ok, removed. Let's hope the problem doesn't recur. —David Eppstein (talk) 20:12, 24 July 2026 (UTC)Reply

See this edit.

—Trappist the monk (talk) 21:30, 25 July 2026 (UTC)Reply

In this case the bad isbn that triggered the edit has been there since the start of the article rather than (as is more usual) being added by another bot later. (It's also an odd choice to include a book review in a list of selected publications but that's not about the bot.) —David Eppstein (talk) 21:50, 25 July 2026 (UTC)Reply

What David said. GIGO. Headbomb {t · c · p · b} 22:00, 25 July 2026 (UTC)Reply