Wikipedia:Bots/Requests for approval/Snotbot 8

The following discussion is an archived debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA. The result of the discussion was

Withdrawn by operator.

Snotbot 8

Operator: Snottywong (talk · contribs)

Time filed: 04:05, Thursday September 8, 2011 (UTC)

Automatic or Manual: Automatic unsupervised

Programming language(s): Python

Source code available: Available on request, uses pywikipedia library with custom code

Function overview: Automatically decline article submissions at WP:AFC which satisfy one of the quick-fail criteria at Wikipedia:WikiProject_Articles_for_creation/Reviewing_instructions#Quick fail criteria

Links to relevant discussions (where appropriate): Wikipedia_talk:WikiProject_Articles_for_creation#Bot_to_enforce_quick-fail_criteria

Edit period(s): Hourly

Estimated number of pages affected: Probably on the order of 10-50 per day.

Exclusion compliant (Y/N): No

Already has a bot flag (Y/N): Yes

Function details: Checks new AFC submissions for the following conditions:

Is the submission completely blank (except for the AfC submission template, and any other default text inserted by the Article Wizard)?
Is the submission in a language other than English? The code will check the text using trigram analysis to determine a confidence factor for if the text is English or non-English. If the confidence factor is sufficiently low, the text is almost certainly in a foreign language.
Are there zero references in the article, zero external links, and no text in the References section (besides the default text entered by the Article Wizard), and the article is not a redirect or dab page?
If any of the above conditions are found to be true, leave a friendly, brief message on the user's talk page warning them that their submission is at risk of being quick-failed.
Once the submission has sat around for at least an hour with no edits made to it, the bot will decline the submission and notify the submitting user as usual.
The bot will also check if the proposed article title already exists. If the proposed article is not a redirect or a dab page, the bot will add a comment to the AfC submission warning that the article title already exists, but the bot will not decline the submission.

Discussion

I don't like the no references bot decline. If the article is on an encyclopedic topic and there are no references, a human should politely interact with the submitter, not a bot. --72.208.2.14 (talk) 03:44, 9 September 2011 (UTC)[reply]

I can tell you that this doesn't happen today. Any AfC reviewer who comes across an article with no references, no external links, and no attempt to add any content to the references section will quick-fail the submission regardless of whether or not it is an encyclopedic topic. Most reviewers will not even read the article in this situation. Furthermore, most AfC reviewers use scripts to help them review articles more quickly, and the scripts simply place a template on the user's talk page to let them know that their submission had been declined. That is the extent of the polite, human interaction that will occur in this typical scenario. This bot request is simply to automate this very predictable situation, in the hopes that the AfC backlog will be cleared more quickly, and the human reviewers will have more time to review serious submissions that actually have a chance of being accepted. —SW— ^babble 03:58, 9 September 2011 (UTC)[reply]

This isn't my experience. Last AfC I submitted where I missed the references due to a copy and paste mangle, I got a polite note on my IP talk page about references, and the reviewer added one. It was a good and appropriate article, a no-brainer, not like a lot of the crap I see submitted there, and having a human editor look at it got the correct ball rolling: a reference and my attention added.

How many articles a day are submitted at AfC? It's not hundreds? And it seems the human editors are already doing the quick denials without a bot for no references. This means, a real article stands a chance of attention. --72.200.106.228 (talk) 16:45, 11 September 2011 (UTC)[reply]

All I'm saying is that the bot's response will be indistinguishable from the typical human response to an article with no references. You will get a template posted to your talk page letting you know that your article has been reviewed, and the article will have a template on it explaining that it needs sources. Your experience described above may have been the exception. As for how many articles are submitted at AfC each day, you can see a history of it at Wikipedia:WikiProject Articles for creation/Submissions#Recent months, and it often tops 100 in a day which is about one article every 10-15 minutes. However, we're expecting that number to increase, perhaps significantly. A new trial will eventually be implemented which will restrict article creation to autoconfirmed users. Non-autoconfirmed users will be directed to AfC (among other places) to get their articles created. We're expecting a large increase in submissions, and not necessarily a correspondingly large increase in reviewers. —SW— ^verbalize 15:35, 12 September 2011 (UTC)[reply]

Your article won't have been reviewed. It will have been scanned by and rejected by a machine. At the least, make accurate templates that state what really occurred: a bot scanned this article and found no references and the bot removed your article from the queue. I sincerely hope folks at AfC are not asking for a bot to template newbie talk pages claiming that a bot scan is a review of a submission. --72.200.106.228 (talk) 08:07, 13 September 2011 (UTC)[reply]

Making variations on the normal talk page template to clarify that it was quick-failed by a bot is a good idea, and definitely possible. If the submitter can't be bothered to follow the extremely clear instructions in the Article Wizard to add at least one reliable source to the article (even an improperly formatted external link), then why should a human reviewer be bothered to waste his/her time reading the rest of the article? —SW— ^talk 14:40, 13 September 2011 (UTC)[reply]

Certain it is that no matter how many policies and guidelines admonish experienced wikipeia editors to be polite to new editors experienced editors can't be bothered to follow these extremely clear admonishments to not bite the newcomers, so, why should newbies be slashed by bots too? Totally do not support this bot as long as it carries a "Bot Bites the Newbie" action. If human reviewers do'nt want to be bothered interacting with new and inexperienced editors, they might find a better place to edit. --68.105.141.221 (talk) 06:52, 14 September 2011 (UTC) (Note that both of these IP's are from the same geographic region, and therefore are probably the same person, despite the appearance of them being two separate people opposing this task.)[reply]

I think we're all getting a little overly emotional here, and I think you would benefit from an education on how things work from the reviewer side of AfC. There is no "Bot Bites the Newbie" action included in this task. In fact, the response that the "newbie" gets would be indistinguishable from the response they would get from the vast majority of human reviewers, except perhaps that they would get the response somewhat faster than usual. Nearly every human AfC reviewer uses a script to help review AfC submissions. At the touch of a button, the script automatically updates the AfC submission page and leaves a templated response on the user's talk page. The bot would do precisely the same thing on articles that have a 0% chance of being accepted by a human reviewer.

On the contrary, no matter how many times new users are reminded and warned by the Article Wizard that their article will be quick-failed if it contains no valid references, many new users choose to ignore these warnings. Why should a human reviewer spend their time reading their submission if the author can't put forth a minimum amount of effort into it? —SW— ^communicate 13:57, 14 September 2011 (UTC)[reply]

Yes, human editors are often rude to newbies. Other crap exists. --68.105.141.221 (talk) 02:20, 16 September 2011 (UTC)[reply]

Could you email me the bot's code for instructive purposes? Thanks, Kudu ^~I/O~ 23:13, 11 September 2011 (UTC)[reply]

This being my 8th bot request, I've now learned my lesson and no longer code tasks before they're approved. So, the code is not yet written. I'd be happy to send it to you when (or if) it does get written though. —SW— ^confess 15:38, 12 September 2011 (UTC)[reply]

As someone starting off at AfC from NPP, this task sounds like a good concept to me. The Blade of the Northern Lights (話して下さい) 20:08, 12 September 2011 (UTC)[reply]

This seems a bit iffy, both from the standpoint that the linked discussion shows a number of concerns being raised and the standpoint that I see some of the same concerns with #2 and #3 and this is an area with large numbers of tasty newbies.

This suggestion is not made as a BAGger, just as a regular editor: The best way to proceed might be to run your bot for a while in a mode that just logs what it would do to a page in your or its userspace (which needs no prior approval, see WP:BOTAPPROVAL), and then give some statistics on how often its proposed action matches what human editors actually did. You might also be able to do the same over old requests (all requests, not just quick-failed ones). If it turns out well, that might serve to ease the concerns being raised. Anomie ⚔ 03:11, 15 September 2011 (UTC)[reply]

Probably not a bad idea. I can make that happen. I'll post here again once I get it done, it'll probably take a few days. —SW— ^gossip 18:21, 15 September 2011 (UTC)[reply]

Good idea. --68.105.141.221 (talk) 02:23, 16 September 2011 (UTC)[reply]

Ok, I set it up and let it run for roughly 20 hours. It found about a few dozen submissions that it would have quick-failed. Note that the code for testing for non-English submissions is not yet finished, so it's not testing articles for that yet. It's also not yet checking if the last edit to the article was made more than an hour ago. The list of articles that it would have quick-failed in this time period (along with links to the specific revision that the bot checked) are at User:Snottywong/Quickfail AfC. Articles failed with a reason of "v" means they didn't have any references, with a reason of "blank" means they were blank submissions. —SW— ^talk 20:13, 16 September 2011 (UTC)[reply]

I let it run some more and added some entries to the table. I checked through many of them and they all looked pretty accurate to me. Any comments? —SW— ^spout 15:46, 19 September 2011 (UTC)[reply]

I just realized that the AfC submission template provides a pretty clear warning when a proposed article already exists (someone told me this before but I didn't see it for myself until now). I've stricken this from the function details above, as it is unnecessary. —SW— ^babble 00:38, 20 September 2011 (UTC)[reply]

Outsider looking in here (aka not in BAG): I used to be pretty active in AfC, and i sort of do feel like the the anon does have a point with the whole no ref thing. Would it be possible to have it post on the IP's page, make a note on the request page, and then check back in an hour? It's been a good while since I've done anything with AfC but it would allow the anons and maybe even a passing by logged in editor a chance to save a potentially good article.Kwsn (Ni!) 02:18, 21 September 2011 (UTC)[reply]

So, do you mean the bot would immediately post a friendly warning on the IP's talk page when it finds an unreferenced article, and then decline the article an hour later if there have been no edits to the article? I kind of like that idea, and I think it's doable. It would require a little bit of extra magic to make sure I don't spam the IP's talk page with multiple warnings about the same article (if, for instance, I stop and restart the bot and it forgets which articles it has already checked), but I don't think that would be too difficult. —SW— ^comment 14:07, 21 September 2011 (UTC)[reply]

Pretty much. It allows users like our anon above who accidentally oops a chance to go back an fix it up before it's declined. Kwsn (Ni!) 14:51, 21 September 2011 (UTC)[reply]

Amended function details above. Thanks for the suggestion. —SW— ^gossip 15:17, 21 September 2011 (UTC)[reply]

Sigh. I forget the depth of utter crap that is uploaded at AfC. I learned years ago to never search the category and look at the rest of it. My objection to robot fails for articles without references still stands. That human editors are just as rude is no justification for a bot to be hostile to newbie editors (yes, other crap exists). The amount of hostility shown to newbie editors on wikipedia grows astronomically as wikipedia grows, and it's disgusting. Let's never add bots to that problem. --68.105.141.221 (talk) 20:32, 21 September 2011 (UTC)[reply]

Your objection is invalid because the bot's activity will be indistinguishable from the typical human reviewer's activity. That you have a problem with the typical human reviewer's activity is immaterial to this bot request (and your opinion is not shared by many others, to my knowledge). If you think the standard templates that are used to notify users of various AfC-related things are rude or hostile, then take that up at Template talk:Afc decline or somewhere similar. You're bringing up your concerns in the wrong venue. If you continue to object to this bot request on the grounds that the notifications it will post on a user's talk page are not acceptable, then your objections will continue to be ignored because the bot's notifications will be identical to the normal notifications used by human reviewers (with the exception of the additional courtesy notification requested by Kwsn above). —SW— ^communicate 21:30, 21 September 2011 (UTC)[reply]

Interesting response, in the wrong way. My objections were not ignored by User:Anomie and User:Kwsn, they were acknowledged. Your refusal to acknowledge my objections and your hostile response to them raises concerns about your ability and willingness to deal with IP editors who have problems with the tags. And that raises another legitimate concern about this bot: that the operator may not be able to deal appropriately with complaints by politely engaging those complaining about bot practices. Yes, there is a difference between a tag placed by a bot and one placed by a human. One can interact with the human. In the case of this bot, interacting with the bot owner may be problematic. --68.105.141.221 (talk) 03:15, 22 September 2011 (UTC)[reply]

Not trying to be hostile, but your objections have been addressed yet you continue to bring them up. Snotbot's talk page is a redirect to my talk page, so any IP who wishes to communicate with a human to get an additional explanation for the rejection of their article is more than welcome to post a message on my talk page. So, now I have shown that this bot task will not change the end result of the AfC submission, will not change the notifications that the user receives, and will not change that user's ability to contact a human with any questions or complaints that they might have. I've also provided an example of dozens of articles that the bot would have quick-failed. Do you have any other objections that haven't already been addressed? —SW— ^converse 14:19, 22 September 2011 (UTC)[reply]

FYI - Striking the language detection from the function details. I'm not able to get it working accurately enough for me to be comfortable with it. There would be too many false positives, and the number of non-English articles submitted is pretty minimal anyway. —SW— ^chatter 18:52, 26 September 2011 (UTC)[reply]

Trial

{{BAGAssistanceNeeded}} Looking for a ruling on this one. Thanks. —SW— ^verbalize 14:06, 26 September 2011 (UTC)[reply]

Approved for trial (20 fails + notifications). Please provide a link to the relevant contributions and/or diffs when the trial is complete. I sympathize with the IP that bots generally appear "ruder" than human editors. However, in this case humans use scripts anyway and the bot would notify the submitter and enter a graceful wait period before failing. (Also, could we see the messages the bot produces before/after fail?) In the end, this would alleviate the work humans need to do anyway and hopefully make them consider other AfCs better. AfC regulars seem to share this standpoint. In the end, I believe this task will be more beneficial than detrimental. Without too many words, I believe this should have at least a trial and subsequent feedback (perhaps a note in the failing messages? I realize this will probably just attract criticism from someone unhappy with their article being failed...). More BAG member opinions welcome. — HELLKNOWZ ▎TALK 19:04, 26 September 2011 (UTC)[reply]

Created a warning template at Template:Afc warning. Putting the finishing touches on the code and will start trialing soon. —SW— ^verbalize 18:40, 28 September 2011 (UTC)[reply]

What about submissions which are accidentally added between the comments? (so that they aren't shown)

What about quickfail-criteria: if fails because of no reference, please check also if URLs included in the text are only: youtube, myspace, facebook, twitter links and (if possible) the title of the submission (.com) and count them as "no reference" since they would be either primary/secondary source or unreliable! mabdul 22:55, 28 September 2011 (UTC)[reply]

The bot doesn't distinguish between text in comments and text that is displayed. If a user creates an article in a comment, the bot will not quick-fail it.

Checking for specific sites would add some complexity to the task, but it is an improvement to consider for future upgrades. It's not something I'll consider adding immediately. Thanks for the input. —SW— ^prattle 23:08, 28 September 2011 (UTC)[reply]

The comment check could be easily implemented by a check of the KB the submission is using... mabdul 23:25, 28 September 2011 (UTC)[reply]

I guess what I'm saying is that I'm not sure it would be desirable for the bot to quick-fail an article if the author accidentally put all the text inside of a comment. —SW— ^confess 00:47, 29 September 2011 (UTC)[reply]

WP:SOFIXIT, for what are we having bots? XD mabdul 02:11, 29 September 2011 (UTC)[reply]

Well, trial is still ongoing. Apparently a few editors conspired to clear out the entire AfC backlog right around when this trial started (not intentionally, of course), and they are keeping the backlog pretty small. Not too many articles are lasting longer than an hour, so there's not a lot for the bot to work with at the moment. The bot has declined 11 articles so far (besides the initial test declines). I'll leave it running all weekend and we'll see if it gets up to 20. As for edits made so far, see the following:

—SW— ^confess 22:36, 30 September 2011 (UTC)[reply]

Nevermind, not running it over the weekend after all. —SW— ^yak 23:34, 30 September 2011 (UTC)[reply]

Trial complete. — see above 3 links for edits made. —SW— ^prattle 14:35, 5 October 2011 (UTC)[reply]

{{BAGAssistanceNeeded}} - AFC backlog has now grown close to 200 submissions, I'd love to be able to trim some of those down. —SW— ^talk 21:19, 12 October 2011 (UTC)[reply]

{{BotDenied}} Sorry, there's no clear consensus on the BRFA or here. The IP has raised valid concerns, which I don't think can be address via this BRFA alone. Please start a discussion on the appropriate forum (e.g. Wikipedia talk:WikiProject Articles for creation, the village pump, etc), in order to establish that there is indeed a consensus and need for such a bot, and that such a bot would not adversely discourage newbies from editing. --Chris 09:05, 14 October 2011 (UTC)[reply]

Reopened per discussion here. I will leave it to another BAG member to decide what needs to be done next. --Chris 11:24, 17 October 2011 (UTC)[reply]

Withdrawn by operator. After more lengthy discussion, it's clear to me that there is no consensus for this task. —SW— ^chatter 16:18, 27 October 2011 (UTC)[reply]

Close as Withdrawn by operator. — HELLKNOWZ ▎TALK 16:49, 27 October 2011 (UTC)[reply]

The above discussion is preserved as an archive of the debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA.