Wikipedia:Bots/Requests for approval/K.Kapil77 Bot

K.Kapil77 Bot

New to bots on Wikipedia? Read these primers!

• Approval process – How this discussion works
• Overview/Policy – What bots are/What they can (or can't) do
• Dictionary – Explains bot-related jargon

• (BRFA
• Approved BRFAs
• talk
• contribs
• count
• SUL
• logs
• page moves
• block log
• rights log
• flag)

Operator: K.Kapil77 (talk · contribs · SUL · edit count · logs · page moves · block log · rights log · ANI search)

Time filed: 08:54, Wednesday, April 28, 2021 (UTC)

Function overview: This Manually-assisted Bot rectifies minor spelling mistakes on less to mid popular pages based on high confidence heuristics.

Automatic, Supervised, or Manual: Manual

Programming language(s): Python

Source code available: Will be made available after approval and testing

Links to relevant discussions (where appropriate):

Edit period(s): daily

Estimated number of pages affected: 10-15 per day

Namespace(s): Articles

Exclusion compliant (Yes/No): Yes

Function details: We have generated a dataset containing minor to larger spelling mistakes using Wikipedia dump based on heuristics like edit distance, popularity, frequency of use. We manually verify about 10-15 such changes everyday on our own designed portal at https://www.iitg.ac.in/cseweb/WikiFeedback. This bot will reflect those changes from our SQL server to Wikipedia bot.

Discussion

Hi there. Interesting project. For the purposes of approval, I'm afraid it would not be responsible for us to approve a closed-source bot, that also requires registration to use, and could violate WP:CONTEXTBOT unless all the edits are supervised. But it appears you might be affiliated with an Indian Institutes of Technology (judging by the domain of your site) so I'm not going to decline this BRFA yet, as your project may be promising, but the function will need tweaking to meet our standards for bots. If we did approve a bot like this we need to have a high level of confidence that it isn't going to make problematic changes (see WP:CONTEXTBOT). You do mention that you'll manually verify 10-15 a day. The bot could be coded as a tool using OAuth to make the edits on a user account directly (see WP:BOTMULTIOP). A bot account isn't usually needed to make a tool like this. All the edits would have to be manually verified by a human before going live on an article to make sure that they're okay. ProcrastinatingReader (talk) 14:05, 28 April 2021 (UTC)

Hello ProcrastinatingReader, Yes I am a research student at an Indian Institutes of Technology. Sorry, I guess it wasn't clear from my previous writing that we have a high confidence dataset which will also be voted upon by many users from our research group on the Portal I mentioned above. Based on User voting (expected 10-15 votes per day per user), our heuristics and recent revisions on similar pages on Wikipedia we would want our bot to make changes, keep tracking if some human user reverts back the changes made and notify us as well as our voters about revisions. I couldn't mention complete heuristics of generating the dataset since it is a live project. Also since all changes are being verified manually, I believe chances of problematic changes is quite low. I can make source code available only after permissions from our group, if its non-availability is a deal-breaker. Please do let me what can be done to get this bot live. K.Kapil77 (talk) 18:29, 28 April 2021 (UTC)

Source code is not required by default to be published, but it's encouraged, and for some tasks it can be preferable to have others look over it.

You can code your algorithm to generate diffs for a change however you like of course (eg using user voting or analysis of revisions). The key for the purposes of this bot is that you need a human to look over the change and make sure it's okay before it actually goes live on an article. If you're looking to move into unsupervised territory, this could maybe be re-evaluated if there's a track record of generated changes being accurate, but in the first instance I don't think a bot could be approved for unsupervised editing in this manner.

As for the 'checking over' part, this is referred to as semi-automated editing. You can use OAuth (see mw:OAuth/For Developers) to directly make the edits on the human editor's Wikipedia account, so it wouldn't necessarily need to go via a bot account and this method wouldn't really require any sort of approval. ProcrastinatingReader (talk) 18:45, 28 April 2021 (UTC)

​ They have already said all edits are manually vetted, so I don't think CONTEXTBOT is applicable. There exist a lot of bots whose edits are manually supervised. Making it an OAuth tool is added complication which is useful only when the creators want to crowdsource vetting of the edits (here they indicate they'll do the vetting themselves). So all in all I don't see any issues here. Having a quick trial will likely help further review. – SD0001 (talk) 12:42, 29 April 2021 (UTC)

Hmm. I guess if these conditions (all edits manually verified by the same editor) will not change in the future it won't actively be a problem. @SD0001: do you want to take it through trial? ProcrastinatingReader (talk) 12:52, 29 April 2021 (UTC)

Approved for trial (25 edits). Please provide a link to the relevant contributions and/or diffs when the trial is complete. Please don't mark edits as minor for the trial (so that they show up in watchlists and RecentChanges). Let us know if you run into any issues. – SD0001 (talk) 13:07, 29 April 2021 (UTC)

Hello! Please consider my edits from User:K.Kapil77 Here : Special:Contributions/K.Kapil77. Most of the edits made on 28, 29 April 2021. No edit has been reverted yet, showing my proof of work. I have made an edit to my sandbox from Bot to show working accuracy of bot code.User:K.Kapil77_Bot/sandbox (I'm using pywikibot)K.Kapil77 (talk) 06:58, 25 May 2021 (UTC)