Wikipedia talk:Bot policy

; Archives
Archive 1	Archive 2
Archive 3	Archive 4
Archive 5	Archive 6
Archive 7	Archive 8
Archive 9	Archive 10
Archive 11	Archive 12
Archive 13	Archive 14
Archive 15	Archive 16
Archive 17	Archive 18
	Archive 19
	Control proposals
	Archive policy
	Archive interwiki (also some approvals for interwiki bots)

Shortcut

WT:B

Request for BAG membership

Per the bot policy, I am making this post to inform the community of my request for BAG membership. Please feel free to ask any questions/comment there. RichardΩ612 Ɣ ɸ 16:09, 5 July 2008 (UTC)[reply]

What is the threshold for manual edits being covered under bot policy?

If someone was creating articles at the average rate of 3 per minute, for periods of nearly an hour at a time, but they claimed to be doing it completely by hand, would their edits still be covered under the Bot policy? For example, 126 stub articles in 47 minutes. Kaldari (talk) 23:58, 22 July 2008 (UTC)[reply]

No. BJ^Talk 00:27, 23 July 2008 (UTC)[reply]

OK, if there is no threshold, I will no longer be requesting BOT approval for any edits that I perform, as I assure you I am going to do them all "manually". Feel free to delete my BOT account. Kaldari (talk) 00:49, 23 July 2008 (UTC)[reply]

… the threshold is automation. BJ^Talk 00:57, 23 July 2008 (UTC)[reply]

So anyone that wants to can circumvent this policy by simply stating that their edits are not automated, no matter how many edits per minute they are making? Kaldari (talk) 01:20, 23 July 2008 (UTC)[reply]

Sir Blofeld is an exceptional case - he really should have a bot flag on his account. But for whatever reason his editing activity is well known and he does not have a bot flag. I would classify his edits as informally approved.

In general, the people who notice "unapproved bots" are the recentchanges patrollers. If one of them notices you making a large number of seemingly automated edits, and you aren't on their mental list of exceptional users, they may leave you a note to find out what's going on. — Carl (CBM · talk) 01:32, 23 July 2008 (UTC)[reply]

Blofeld requested a bot account to create articles and it was denied when it was determined he would still be individually creating each article and it would not be automated in any way. MBisanz ^talk 01:40, 23 July 2008 (UTC)[reply]

If that is true, doesn't Wikipedia need some way to deal with regulating prolific manual editing as well? In case no one has noticed, Sir Blofeld has created 30,000 of the shortest stubs on Wikipedia, many without any sources or references, and some with questionable sources (like maplandia). I believe Wikipedia would be better off without these stubs until enough information is available on the internet to create decent articles, whether by bot or manually. But apparently the rest of Wikipedia has no problem letting him run rampant with no oversight whatsoever. I don't see how this is improving Wikipedia. Kaldari (talk) 03:17, 23 July 2008 (UTC)[reply]

If someone (such as myself) objects to the mass editing habits of someone who is "informally approved", what then is the appropriate channel for raising that objection? I already tried raising the objection on their talk page, but everyone there insists that there is no policy or guideline against what they are doing. Am I left then to manually review this user's edits as fast as they make them, nominating every article I can't verify for deletion? That would obviously be impossible. Do I have any alternative other than just completely ignoring the fact that Wikipedia is being flooded with thousands of unverifiable stubs (a large percentage of which seem to be based on single entries in GEOnames or maplandia)? Kaldari (talk) 03:23, 23 July 2008 (UTC)[reply]

WP:VPM is one place you could start a discussion if you disagree with the articles being created. If your concern is with sourcing, you could ask on WT:V. If your concern is with notability, WT:N. Have you tried raising your concerns directly with Blofeld? — Carl (CBM · talk) 03:48, 23 July 2008 (UTC)[reply]

Please see Wikipedia:Village pump (proposals)/FritzpollBot for some discussion about article creation. Making regular edits at over 5/minute for an extended time is likely to be considered bot-like, but I don't think we've ever discussed if semi-automated article creation at a somewhat lower rate would be enough to be bot-like. But you could always send the stubs to articles for deletion, perhaps in groups, or possibly to speedy deletion in some cases. Gimmetrow 03:57, 23 July 2008 (UTC)[reply]

Well, please don't mass-tag them. You can make a list and use a single AFD for the entire list. I don't believe there are any CSD criteria that apply to stubs of geographic locations. — Carl (CBM · talk) 04:02, 23 July 2008 (UTC)[reply]

I would hardly call working for 45 minutes at a time "bot-like." If he was working for several hours at a time, then maybe. But even semi-automatic scripts aren't even covered under the bot policy. As long as he's manually approving each edit, there's no real problem, at least as far as compliance with the bot policy goes. I fail to see how "no oversight whatsoever" is a huge problem - besides the usual recentchanges and newpages patrol, there's as much oversight as there is for any other user. Bots generally have even less oversight after approval, so I'm not sure how getting a bot flag would help with that. If the rest of Wikipedia has no problem with this, that's generally called a consensus Mr.Z-man 18:24, 26 July 2008 (UTC)[reply]

English Wikipedia opt into the global bot policy for interwiki bots

On Meta there is now a global bot policy for interwiki bots. The new policy will allow interwiki bots to have a global bot flag on wikis that allow it. I propose we opt into the system, while I don't think it was created with en.wp in mind, it would still be a great benefit to us. The current process for approving interwiki bots is woefully lacking. Now the bots are approved with only a short sanity trial without regards if the interwiki links made are even correct. Allowing the bots to be globally approved, with input by native speakers and accountability across many wikis would be a large improvement over the rubber stamping system we have now. BJ^Talk 18:03, 26 July 2008 (UTC)[reply]

I'd support the idea, but it is important to clarify that these bots are to be used solely for interwiki links, and that the flag is granted by stewards. . - Rjd0060 (talk) 18:08, 26 July 2008 (UTC)[reply]
I would definitely support this. Interwiki bots that already operate in multiple projects are pretty much guaranteed approval here. This would just make it easier for the operators. Mr.Z-man 18:33, 26 July 2008 (UTC)[reply]
I would support the idea. Right now it seems as if there is a major language barrier between wikis allowing interwiki bot operators to get away with a lot more than they'd be able to normally. --Nn123645 (talk) 19:14, 26 July 2008 (UTC)[reply]
My only concern is that we don't have control over what Meta is to include later in their "Global bots" section. So as long as we clearly define what global bots are allowed to do locally (i.e., only interlanguage work), it sounds fine to me. --MZMcBride (talk) 19:52, 26 July 2008 (UTC)[reply]
Sounds fine to me too, with the caveat mentioned by MZMcBride above. —Ilmari Karonen (talk) 19:56, 26 July 2008 (UTC)[reply]
I'd agree on the merits of opt-ing in for interwiki bots only. Global bots found to be doing other work without local approval should be blocked with judicious use of the banhammer. Q ^{T C} 05:18, 27 July 2008 (UTC)[reply]
The addition of EnWP to the standard bot policy list was reverted. I support our addition per BJ. —Giggy 05:35, 27 July 2008 (UTC)[reply]
I support adding the clear local definition here and at our global rights policy: WP:GRU. MBisanz ^talk 02:10, 29 July 2008 (UTC)[reply]
Comment:Now the bots are approved with only a short sanity trial without regards if the interwiki links made are even correct.? Are you saying that's the current system, or what it will be? Because that's not how interwiki bots are approved here. So, does the global bot policy have any consequences for us other than default approval for a few interwiki bots? Global-approved interwiki bots could presumably still be blocked locally if necessary. Also, even under the global system, bots would still need to get individual approval on a few wikis first, so we'll still review some. Gimmetrow 02:21, 29 July 2008 (UTC)[reply]
That is how they currently are approved. The operators often speak poor English and the BAG (unless I'm missing something here) can't read the language of the interwiki links to see if they are correct. There is almost never any discussion about the bot, they are quickly trialled and approved. BJ^Talk 02:44, 29 July 2008 (UTC)[reply]
Well I guess I'm the ogre then. While I assume the linking algorithm is "correct" in established frameworks, I do check what wikis the bot/operator has sought or received approval and spot-check a couple links in a language I know. Gimmetrow 02:52, 29 July 2008 (UTC)[reply]
Comment: So may enwiki be added now to the list of opt-inned wikis? — VasilievV ² 13:06, 2 August 2008 (UTC)[reply]
I'd say so. —Giggy 13:38, 2 August 2008 (UTC)[reply]
I'd support, but let's not make the mistake we always make and assume that "support on some bot talk page" == "support from community" when community has not been informed. Any objections to refactoring this as an evil poll and spamming some boards with it? Happy‑melon 13:55, 2 August 2008 (UTC)[reply]
Go for it, also I've worked what I hope is the spirit of this discussion into our WP:GRU policy. MBisanz ^talk 20:13, 2 August 2008 (UTC)[reply]

Above comments refactored from a threaded discussion. Happy‑melon 20:53, 2 August 2008 (UTC)[reply]

This is after all where we go to make decisions about bots. Chillum 21:46, 2 August 2008 (UTC)[reply]

Support Of course. Chillum 21:04, 2 August 2008 (UTC)[reply]
Support Prodego ^talk 21:44, 2 August 2008 (UTC)[reply]
Support -- Yes, it would speed up approvals. Meta is the best place to handle interwikibots. =Nichalp «Talk»= 06:08, 3 August 2008 (UTC)[reply]

I have made the changes to the relevant Meta pages: [1][2] —Giggy 06:20, 3 August 2008 (UTC)[reply]

It looks like you want to opt in to global bots and otherwise not use the standard bot policy. That is possible, but the requirements for global bots are defined by that policy rather than en-Wikipedia policy. (Of course, you can block global bots that operate here against the local policy.) —{admin} Pathoschild 22:35:26, 03 August 2008 (UTC)

That sounds about right, I think. —Giggy 01:29, 4 August 2008 (UTC)[reply]

BAG membership nomination

Per the bot policy, I am making this post to inform the community of a request for BAG membership. Please feel free to ask any questions/comment there. SQL ^{Query me!} 03:14, 28 July 2008 (UTC)[reply]

Request for corporate memory on common bot screwups

I have blocked Lightbot for exceeding its mandate and subsequently screwing up. It was converting raw units to {{convert}} calls e.g. "25 ft" -> "{{convert|25|ft}}". I first noticed the problem when it did so in a direct quote,[3] and shortly afterwards someone pointed out that it had twice made such a conversion to a book title.[4][5]

The reason I am bringing this here is because we seem to be going over the same ground again and again. It has long been established that bots that make textual changes to articles don't work, because it is impossible to distinguish between material that is amenable to improvement, and material that must be preserved exactly as is, such as quotes, book titles, etc. But we don't seem to be learning from our mistakes. It frustrates me that we have to keep going over the same stuff, and I'm sure it must frustrate the BAG even more. Don't we have some kind of corporate memory in this area? e.g. required reading for bot operators? Somewhere where we can stick a bloody great blinking text box that says "Bots must not make textual changes!"

Hesperian 00:21, 1 August 2008 (UTC)[reply]

The project namespace is supposed to be our coporate memory (duh), but it has been hijacked, and is currently running one of the biggest and longest running nomics in the history of mankind. :-P --Kim Bruning (talk) 00:30, 1 August 2008 (UTC)[reply]

I have to say, as an outsider to this discussion, but as someone with plenty of experience writing bots and web crawlers in non-Wikipedia contexts, I'm not very clear what is meant by "bots must not make textual changes". What I assume you mean is that a bot mustn't treat an article as an opaque string and (effectively, if not actually) apply a regex to it. A bot must be able to parse the logical structure of the article, and act on that. Presumably implied in this is a requirement that bots should only alter parts of a article that are sufficiently structured that the changes of a false positive are practically zero.

So, for example, if I have a bot that replaces º, the Spanish masculine ordinal indicator, with °, the degree symbol. When would it be reasonable to make this replacement? If an article contains the sentence "Longyearbyen's latitude is 78ºN" would it be appropriate for a bot to apply a regex such as s/\<([0-9]\+)º([NSEW])\>/\1°\2/g that would replace them? I can't at present imagine a situation where an incorrect conversion might arise, but there may well be such situations. But I can't see what "textual change" might mean if this is not an example of one. However, I think we have quite a lot of bots that make precisely this sort of change.

— ras52 (talk) 12:44, 1 August 2008 (UTC)[reply]

Copy-paste the text of a page (as served, not as it appears in the edit box) into a raw text editor, such as Notepad or vi. The result is the text, with no links, no typeface, no layout, no images. That is what I mean by The Text.

Every time a bot tries to make wholesale changes to The Text, it screws up. We've had bots that capitalised florida in specific epithets. We've had bots that "corrected" the spelling of the French word mariage. We've had bots that fixed spelling in direct quotes, and now we have a bot that puts unit conversions in book titles.

You've put forward the safest example you can find, but even in this case it is easy to construct a example where the bot would fail. What would your bot do to the following sentence?

As an example of a textual error that a bot could safely fix, Ras52 offered the sentence "Longyearbyen's latitude is 78ºN".

Hesperian 13:19, 1 August 2008 (UTC)[reply]

The bot would change it. An editor then notices the mistake, reverts the edit and notifies the botop who blacklists the article or changes the coding so that anything inside quotemarks are ignored. The bot then goes back to making 499 good edits for every 1 obscure mistake. A question: do the edits ClueBot make fit your definition of a textual change? ~ AmeIiorate _{U T C} @ 14:04, 1 August 2008 (UTC)[reply]

ClueBot is a special case and doesn't mark its edits +b. But Lightbot was approved as manually assisted, so the operator should be looking at context. BJ^Talk 14:09, 1 August 2008 (UTC)[reply]

Where did this discussion go wrong? This thread was not intended to start a competition to see who can think of the most foolproof example of a bot-fixable error, nor a competition to see who can trip me up and make my definitions/terminology look silly.

My point is that people are repeatedly letting loose bots that change text erroneously, because the bot-owners fail to realise that some text, such as quotes and names, should not be "corrected". The fact that we are eternally going over the same ground, and making the same mistakes, instead of learning from them, is a problem. What can we do about it?

Hesperian 14:21, 1 August 2008 (UTC)[reply]

I'm sorry if you thought I was trying to trip you up and make your definitions/terminology look silly. But I'm still genuinely struggling to understand what you mean. Do you think that it is reasonable for a bot to change 78ºN to 78°N irrespective of where it is? I'm assuming that you don't. But if so, can you give some examples of edits that you think a bot could legitimately make? — ras52 (talk) 14:56, 1 August 2008 (UTC)[reply]