Wikipedia:Bots/Requests for approval/Faebot
- 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Withdrawn by operator.
Operator: Fæ (talk · contribs · SUL · edit count · logs · page moves · block log · rights log · ANI search)
Time filed: 16:56, Sunday March 16, 2014 (UTC)
Automatic, Supervised, or Manual: Supervised
Programming language(s): Python
Source code available: Pywikipedia
Code to be written during test period and can be published on the associated wmflabs account or wherever is most convenient.
Function overview: Creation of 900 start class articles about Site of Special Scientific Interests in Wales.
This project is by request of Llywelyn2000 (talk · contribs), a bureaucrat on cy.wp, who has set up similar articles about Site of Special Scientific Interests on the Welsh Wikipedia and has asked me to look at his database to use same data to create articles in English. An example is at Llanbadrig - Dinas Gynfor. This is part of a Wikimedia UK long term supported project in Wales, though I am acting as an unpaid and independent volunteer.
One outcome I would like for this limited task is to establish Faebot for future limited run projects, similar to the scope at Commons:User:Faebot. Potentially projects might be related to my media projects for Wikimedia Commons, but I do not currently envisage any maintenance tasks, so these will be projects with a short life and measurable outcomes to be reported.
Faebot has a good track record of projects on Commons by following an unhurried passive tests/re-writes, beta tests/consultation, deployment, review and notification process. Over 2.5m edits have been made,[1] the largest single project being place categorization for Geograph images (the main part of which has taken 18 months).
Sample runs: User:Fæ/sandbox#Wales_SSSI
Links to relevant discussions (where appropriate): Wikipedia_talk:Welsh_Wikipedians'_notice_board#SSSIs_in_Wales - I would expect feedback (allowing at least a week, more if issues arise) during my test run at this noticeboard.
Edit period(s): One time run
Estimated number of pages affected: 900
Exclusion compliant (Yes/No): Yes
Already has a bot flag (Yes/No): No
Function details: Process data for 900 SSSI sites, validate against external sources as needed, create articles.
Optionally, maintain a report of status of the 900 articles to support Wikiproject Wales, though this may be achieved using other tools.
Discussion
[edit]- Note: Faebot is indefinitely blocked right now (log), per an ArbCom remedy that restricts Fæ to using a single account. Sven Manguard Wha? 05:52, 17 March 2014 (UTC)[reply]
- If I have a legitimate project with support from the community that would sensibly be run from my bot account rather than doing precisely the same thing but "hidden" under my Fæ account, this would be a good rationale to go back to arbcom for a waiver; Arbcom are not insensitive to good faith content creation projects and I am well established as a non-controversial bot-writer. --Fæ (talk) 07:42, 17 March 2014 (UTC)[reply]
- I expect you already know this, Fæ, but it bears restating for anyone else watching: That would need to be done before the bot could be unblocked, of course. Having consensus here might help the ArbCom debate, OTOH ArbCom could make it conditional on "if BAG approves the request". I don't know which they'd prefer, so I'll leave it up to you to decide when to request it. Anomie⚔ 11:37, 17 March 2014 (UTC)[reply]
- Yes, thanks. I would rather get evidence of a reasonable consensus before taking it up, overall I think that works rather better. If there is no interest at BAG, then there is no point worrying about it and I can move on to other stuff.
--Fæ (talk) 12:19, 17 March 2014 (UTC)[reply]
- Yes, thanks. I would rather get evidence of a reasonable consensus before taking it up, overall I think that works rather better. If there is no interest at BAG, then there is no point worrying about it and I can move on to other stuff.
- I expect you already know this, Fæ, but it bears restating for anyone else watching: That would need to be done before the bot could be unblocked, of course. Having consensus here might help the ArbCom debate, OTOH ArbCom could make it conditional on "if BAG approves the request". I don't know which they'd prefer, so I'll leave it up to you to decide when to request it. Anomie⚔ 11:37, 17 March 2014 (UTC)[reply]
- If I have a legitimate project with support from the community that would sensibly be run from my bot account rather than doing precisely the same thing but "hidden" under my Fæ account, this would be a good rationale to go back to arbcom for a waiver; Arbcom are not insensitive to good faith content creation projects and I am well established as a non-controversial bot-writer. --Fæ (talk) 07:42, 17 March 2014 (UTC)[reply]
- The above note aside, I feel that bot-created articles are something that Wikipedia needs to take a strong stance against. Just about every automated/semi-automated, batch created article that I know of is a few lines long, lacks depth, and never gets edited by a human afterwards (so it never gets improved from being a very short stub). Wikipedia doesn't need 900 new stubs, no matter how notable the subject. I'd rather have something not covered at all than have it covered at such a poor level of quality and completeness that it's embarrassing to show people. Sven Manguard Wha? 06:39, 17 March 2014 (UTC)[reply]
- The aim is for start class articles rather than stubs. The Wikimedia UK project this is a part of intends to enable public participation to simultaneously gain new editors for the English Wikipedia, Welsh Wikipedia and other languages. Without this basic framework of articles about SSSIs for the public to find (on their mobile devices) as they ramble around Wales, there are even higher barriers to entry. I am aware that bot-created articles remain controversial, I am not naturally in favour of them myself in many cases, probably making me the sort of cautious bot-writer that you would want to deploy these and only inclined to spend my time on a topic of high public value. Being part of an improvement programme and with careful testing and set-up, there is little new here that has not been successful in the past and in some cases specially funded by the WMF and lauded as successful initiatives in the press. --Fæ (talk) 07:42, 17 March 2014 (UTC)[reply]
Needs wider discussion.: Per WP:MASSCREATION, this should be advertised at WP:VPR and the talk pages of relevant WikiProjects (e.g. whichever projects would have their banners put on the talk page). Thanks. Anomie⚔ 11:37, 17 March 2014 (UTC)[reply]
- Good suggestion. I'll engage WikiProjects if I get as far as producing a test sample to discuss and I am sure that the Welsh team can help me with appropriate notices after that. If it is still controversial, I can even keep a test set sandboxed for a wider preliminary discussion beyond BAG and WikiProject Wales. I have no reason to hurry this, though in the longer term there is a project that would like to be able to rely on these. --Fæ (talk) 12:19, 17 March 2014 (UTC)[reply]
- @Sven Manguard: Does your "embarrassing to show people" comment refer to the examples Fæ has provided, or is it hypothetical? Having examined them, I see no reason why I would be embarrassed to show them to people; indeed, I would find them a useful tool to engage new editors, in their expansion and illustration. Nor would I feel the need to tag, PROD or AFD them if created by a human. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 12:20, 14 June 2014 (UTC)[reply]
- It's not hypothetical, we've seen bot created articles before, and they're tiny, worded awkwardly, have a low number of sources, and never get expanded (because, as a general statement, if no one cared to create articles on the subjects eleven years in, no one is going to start caring). Looking at the examples at [2], all of the complaints I normally have about bot generated articles hold true. Simply put, an 80 word, 500 character (400 without spaces), is too small to be useful to readers. At that point, it's better to let readers find something better, eleswhere. Sven Manguard Wha? 16:32, 14 June 2014 (UTC)[reply]
- Good grief, what a wholly negative comment to be faced with. I find this depressingly discouraging in terms of thinking of investing my time helping the Wales projects with more programming effort. Sven Manguard, linking to versions of pages I have *just* removed as evidence is unhelpful, you must have known I did that for a reason and nobody could miss the huge orange notice I put on the page. I was going to spend a good chunk of my Sunday taking a second look at this, however I think I'll just tidy the garden instead, as a more realistically productive investment of time that I might actually get some thanks for. --Fæ (talk) 17:31, 14 June 2014 (UTC)[reply]
- You're right, gardening would be a more productive investment of your time. I don't think you'll ever get a version out that won't be crap. That's nothing against you personally, it's just that everything that I have ever seen has shown that what you are trying to do simply can't be done. Wikipedia has tens of thousands of bot-created articles, and except for a tiny number that were later heavily improved by a human, they are all next to useless. If you think that you can whip up a bot-created, 2000 character, half-dozen source article that has sentence flow and doesn't read like it was made using stencils, go for it. Anything less than that, and you're just adding junk into a project that has far, far too much of that already. Until you show me that it can be done at that level, I will continue to believe that bot-created articles are harmful to the project, and that the practice should be banned. Sven Manguard Wha? 19:59, 14 June 2014 (UTC)[reply]
- You seem to be putting the bar higher for an article created by a bot than for a human. I can see no justification for this. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:41, 15 June 2014 (UTC)[reply]
- No, not really. I would just as readily enforce those standards on humans, and delete any article that fails to meet them. Wikipedia is far too accepting of garbage articles, however, so that's not a fight worth pursuing. Sven Manguard Wha? 19:20, 15 June 2014 (UTC)cause[reply]
- In which case, your objections are not substantial, because they are not based on current policies, practices or community norms. If this (or a similar, future) proposal meets those, then it should pass. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 19:31, 15 June 2014 (UTC)[reply]
- No, not really. I would just as readily enforce those standards on humans, and delete any article that fails to meet them. Wikipedia is far too accepting of garbage articles, however, so that's not a fight worth pursuing. Sven Manguard Wha? 19:20, 15 June 2014 (UTC)cause[reply]
- You seem to be putting the bar higher for an article created by a bot than for a human. I can see no justification for this. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 18:41, 15 June 2014 (UTC)[reply]
- You're right, gardening would be a more productive investment of your time. I don't think you'll ever get a version out that won't be crap. That's nothing against you personally, it's just that everything that I have ever seen has shown that what you are trying to do simply can't be done. Wikipedia has tens of thousands of bot-created articles, and except for a tiny number that were later heavily improved by a human, they are all next to useless. If you think that you can whip up a bot-created, 2000 character, half-dozen source article that has sentence flow and doesn't read like it was made using stencils, go for it. Anything less than that, and you're just adding junk into a project that has far, far too much of that already. Until you show me that it can be done at that level, I will continue to believe that bot-created articles are harmful to the project, and that the practice should be banned. Sven Manguard Wha? 19:59, 14 June 2014 (UTC)[reply]
- Good grief, what a wholly negative comment to be faced with. I find this depressingly discouraging in terms of thinking of investing my time helping the Wales projects with more programming effort. Sven Manguard, linking to versions of pages I have *just* removed as evidence is unhelpful, you must have known I did that for a reason and nobody could miss the huge orange notice I put on the page. I was going to spend a good chunk of my Sunday taking a second look at this, however I think I'll just tidy the garden instead, as a more realistically productive investment of time that I might actually get some thanks for. --Fæ (talk) 17:31, 14 June 2014 (UTC)[reply]
- It's not hypothetical, we've seen bot created articles before, and they're tiny, worded awkwardly, have a low number of sources, and never get expanded (because, as a general statement, if no one cared to create articles on the subjects eleven years in, no one is going to start caring). Looking at the examples at [2], all of the complaints I normally have about bot generated articles hold true. Simply put, an 80 word, 500 character (400 without spaces), is too small to be useful to readers. At that point, it's better to let readers find something better, eleswhere. Sven Manguard Wha? 16:32, 14 June 2014 (UTC)[reply]
- The aim is for start class articles rather than stubs. The Wikimedia UK project this is a part of intends to enable public participation to simultaneously gain new editors for the English Wikipedia, Welsh Wikipedia and other languages. Without this basic framework of articles about SSSIs for the public to find (on their mobile devices) as they ramble around Wales, there are even higher barriers to entry. I am aware that bot-created articles remain controversial, I am not naturally in favour of them myself in many cases, probably making me the sort of cautious bot-writer that you would want to deploy these and only inclined to spend my time on a topic of high public value. Being part of an improvement programme and with careful testing and set-up, there is little new here that has not been successful in the past and in some cases specially funded by the WMF and lauded as successful initiatives in the press. --Fæ (talk) 07:42, 17 March 2014 (UTC)[reply]
- No. No articles should ever be created from databases by bots. When someone can find multiple reliable sources that discuss each of these places in detail, then an article can be created which provides information from these multiple sources and provides citations to them. These atlas bots need to stop.—Kww(talk) 00:53, 12 May 2014 (UTC)[reply]
- Also recalling the last time that an editor created a great many articles from database entries and the level of pain that volunteers went through in attempting to clean up the database formula created articles, I think exercising caution is a wise precedent. Hasteur (talk) 14:37, 16 May 2014 (UTC)[reply]
- Can you point to problems with the sample articles Fæ has provided, rather than with edits created by another bot operator, using a different set of data, and different code? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 12:16, 14 June 2014 (UTC)[reply]
- I support this proposal. While unregulated mass-creation of articles by bots may be a problem, this is a relatively small, clearly-defined set of notable subjects. Perhaps people would be reassured if Fæ ran the project in small batches (say 50 articles/day) and be stopped if issues arise; with the limit gradually increasing, otherwise. The sample articles at User:Fæ/sandbox#Wales_SSSI look fine, and are a benefit to our project. That is the primary issue. Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 12:03, 14 June 2014 (UTC)[reply]
Re-proposal
[edit]I would like to re-propose this as to create and maintain a burn-down list of locations based on Cadw's database of Listed Buildings in Wales (which I have been supplied by email). This would be a simple set of working tables, probably a sub-set under a WikiProject which can be transcluded into discussion. The burn-down part of this would be based on tracking which articles had been created on a daily basis. It may be possible to use some WikiData identities, however as these are not yet fully available for buildings, this may have to be developed in parallel, or subsequently.
If anyone would prefer the current proposal to be closed and this created as a new proposal, I would be happy to re-organize this page. --Fæ (talk) 11:40, 12 May 2014 (UTC)[reply]
- Fæ Just a thought (or a rube golberg machine), but what about having the script generate the progress, then you updating the progress table by hand. You say that the updates will happen once a day, so that isn't too bad in terms of an action. Heck, if you post only to your userspace you may not even need to request BRFA. Just some ideas. Hasteur (talk) 14:37, 16 May 2014 (UTC)[reply]
- I am very cautious about this sort of approach. It in effect disguises bot generated edits, one even runs the risk of encouraging automatic article creation by the back-door. Rather than later being accused of manipulating policy or working around a consensus, I would rather be as transparent as possible by having the table/backlog clearly generated by a bot account and publishing the source code for it.
- I have no particular deadlines for this, but some of the Welsh volunteers are agitating for this to get underway as part of projects this summer. If there is a quicker process, perhaps by the BAG accepting this going on with a couple of backlog pages in talk space for an initial test period before full approval, then I would like this stated here as an interim consensus and I'll get on with it. As well as creating the script I need to notify Arbcom in order to have Faebot active again for this purpose. --Fæ (talk) 14:50, 16 May 2014 (UTC)[reply]
Fæ I think that by the time we are discussing this, at least 100 pages could already have been created semi-manually. -- Magioladitis (talk) 22:33, 23 June 2014 (UTC)[reply]
User:Anomie any opinions on the re-proposal? -- Magioladitis (talk) 07:07, 1 July 2014 (UTC)[reply]
- Just the same as was said before: let's see a village pump discussion where people support this. Anomie⚔ 10:41, 1 July 2014 (UTC)[reply]
- Fæ, Anomie I posted on Village pump asking for feedback. -- Magioladitis (talk) 10:46, 1 July 2014 (UTC)[reply]
Wikipedia:Village_pump_(miscellaneous)/Archive_46#Bot_to_create_start_class_articles_about_Site_of_Special_Scientific_Interests_in_Wales have nothing. 9 days, no comments and discussion now in archive.
Fæ, Anomie what do you suggest to be done? -- Magioladitis (talk) 09:52, 10 July 2014 (UTC)[reply]
- I have found this 4 months long process unpleasantly hostile, making me feel like I was asking to steal something, or was setting out to deliberately disrupt the encyclopaedia (see comments above about "embarrassing to show people", "garbage articles" and this discussion being "a fight"). I was approached in good faith to look at this, it was not even my original initiative. If I am going to help the English Wikipedia in the future with small amounts of helpful automation (this example being a tiny project compared to my highly respected bot related work on Commons), then I do not want to use my volunteer time having to defend my good faith work, though I am always happy to take on feedback and adapt my approach.
- I now consider this proposal indefinitely parked. If someone wants to spend their time coming up with a proposal that is warmly supported by the wider community, I'll look at helping again if I am asked nicely in a few months time.
- In the meantime, I'll go back to focus on Commons with 250,000 historic images of the U.S. and helping the Wellcome with another 100,000 high quality medical related images that will no doubt be far more use to the Wikipedia and a much more productive use of time in the coming month. I tend to get mostly supportive comments and buckets of good faith from the community there when we discuss how best to use bots. --Fæ (talk) 11:52, 10 July 2014 (UTC)[reply]
Fæ I left a comment at this request immediately after I joined the BAG. I gave 10 days to the Village pump to receive any comments. I am willing to give you the opportunity to create 30 pages as a bot trial. Do you want it? I think it would be useful and will attract more eyes here. The edit summary must contain a link to this page. Thanks. -- Magioladitis (talk) 13:58, 10 July 2014 (UTC)[reply]
- I would have looked at this a month ago, but after so long I have stopped believing this proposal is worthwhile compared to the volunteer time it would take defending it. I'm busy doing lots of good stuff in preparation for Wikimania, which is more productive for my time. I suggest you contact Wikimedia UK as there appear to be bot writing chapter members available and happy to help with these sorts of chapter based projects. --Fæ (talk) 14:55, 10 July 2014 (UTC)[reply]
Withdrawn by operator. -- Magioladitis (talk) 13:37, 11 July 2014 (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.