User talk:ClueBot Commons/Archives/2017/April


Regarding Chitpavan edit

Dear sir , The edit which you reverted was not vandalism but correction on my part.The paricular article is very political and contains biased views about the community.The addition which I did is quoted from a scholar Ravinder kumar whose source can be found in the same article itself (no. 29 ). I request you to please restore the changes Abcscholar (talk) 07:23, 23 March 2017 (UTC)

Abcscholar Please read the bright red notice at the top of the talk page--5 albert square (talk) 07:41, 23 March 2017 (UTC)

#Hamiltrash vandalism

There is a reasonably well-defined category of persistent vandalism going on, related to the musical Hamilton, mostly consisting of inserting quotes from the lyrics into related articles. Can the bot take a special pass to review and learn from manually-reverted vandalism, on a dataset consisting (for example) of the following pages?

  1. pages that link to Alexander Hamilton
  2. pages that link to Aaron Burr

That's broader than necessary, so please let me know if you'd want a narrower dataset of examples. Thanks! Lwarrenwiki (talk) 14:15, 23 March 2017 (UTC)

False positive

I cannot submit a false positive through the other link. It appears to be broken. This edit is clearly not vandalism. Anyone who can access the link please submit this. Revert ID: 2995778. Thank you! RoCo(talk) 01:08, 5 April 2017 (UTC)

2996586 was also a false positive. Nate (chatter) 15:50, 5 April 2017 (UTC)

Also this: [1] was a false positive

This is also a false positive 2996528 Koden02 (talk) 05:29, 8 April 2017 (UTC)

Another false positive

I tried to report a false positive (revert ID 2999792) on behalf of a new user, but the link from User:ClueBot NG/FalsePositives delivered up a 404 error. Pinging User:Rich Smith and User:DamianZaremba. RivertorchFIREWATER 12:56, 8 April 2017 (UTC)


  Done Though archives are not supposed to be edited, I'm taking an exception here and putting this so there won't be duplicate reports. Reported all false positives above. RoCo(talk) 05:49, 21 April 2017 (UTC)

A barnstar for you!

  The Anti-Vandalism Barnstar
Amazing work! I see you every day. Great work! I Am Chaos (talk) 14:01, 7 April 2017 (UTC)

3 million reverts by CBNG!

Congratulations to ClueBot NG for accumulating 3 million reverts! Great job ClueBot NG. We appreciate all the anti-vandalism work that you do for us. Yoshi24517Chat Very Busy 19:46, 9 April 2017 (UTC)

Yes! Thank to CBNG and team for lowering our workload and creating a feed that allows us to sort edits. L3X1 (distant write) 20:41, 10 April 2017 (UTC)

Suggestion

Black Kite suggested adding "gonzo" to ClueBot's list of things to look for, per the discussion about an IP hopping vandal found here. Basically this user adds the words "gonzo" and "trump 2020" to articles in an effort to troll. EvergreenFir (talk) 22:48, 10 April 2017 (UTC)

@EvergreenFir: CBNG doesn't use a list of words to detect vandalism, it uses a database of hand-picked edits to learn what is and isn't vandalism. For things like that an edit filter might work better. —k6ka 🍁 (Talk · Contributions) 16:32, 11 April 2017 (UTC)
@K6ka: Thank you! EvergreenFir (talk) 18:35, 11 April 2017 (UTC)

A question.

It seems that Cluebot III and Cluebot NG are slowly falling into code-disrepair. As Wikipedia changes, so does vandalism. Meanwhile, Cluebot NG isn't detecting 40% (Roughly) of vandalism. Yes, Cluebot NG makes our lives easier, but i think it's time for a rewrite. Also, Cluebot III is just... Yeah. Maybe it's time to deactivate Cluebot III and create a new, fully updated version of it similar to how the original Cluebot got deactivated in favor of Cluebot NG. Terrariola (talk) 10:45, 6 April 2017 (UTC)

@Terrariola: ClueBot NG doesn't need a code rewrite, just needs to have its database of edits updated for recent vandalism trends. CBNG was designed to learn and adapt over time.
As for CBIII... there's lowercase sigmabot III, which has stood the test of time well. —k6ka 🍁 (Talk · Contributions) 12:45, 6 April 2017 (UTC)
I agree both the bots need improvements, but sadly the 'life' things gets in the way via consumption of large amounts of time. All the code is opensource on GitHub and I'd be happy to review any suggested changes. Moving the report interface to use oauth and bringing the review interface back is long over due. - Damian Zaremba (talkcontribs) 20:18, 14 April 2017 (UTC)

ClueBot III

Out of curiosity, is there any reason that ClueBot III isn't running? Elisfkc (talk) 16:04, 11 April 2017 (UTC)

Apologies, it looks like the process stopped and the watcher failed to restart it. It should be back again now - Damian Zaremba (talkcontribs) 20:15, 14 April 2017 (UTC)

Yoga as exercise or alternative medicine

You archived that page ([2]) but the page has moved and no button to access that archive is on that "new" talk page. — Preceding unsigned comment added by 68.151.25.115 (talk) 07:47, 17 April 2017 (UTC)

ClueBot report interface...

...is down. ««« SOME GADGET GEEK »»» (talk) 05:03, 3 April 2017 (UTC)

As are the IRC feeds. West.andrew.g (talk) 17:14, 3 April 2017 (UTC)
And oddly enough, CBNG itself is running. —k6ka 🍁 (Talk · Contributions) 12:15, 4 April 2017 (UTC)
CB III (archiving), on the other hand, is down. ~ Rob13Talk 12:20, 4 April 2017 (UTC)

IRC is still down. I will ping the operators. Thanks, West.andrew.g (talk) 13:37, 6 April 2017 (UTC)

After a couple day fix, IRC is down again. West.andrew.g (talk) 14:10, 10 April 2017 (UTC)
... and back again. I'm going to put the old cronjobs back in place as it seems bigbrother is not great at restarting things when needed. Apologies for the interruption, I was away with a laptop that was lacking the SSH key to fix this. - Damian Zaremba (talkcontribs) 20:20, 14 April 2017 (UTC)

All the while ClueBot NG's false positive report page is still not up. Get ready to handle lots of false positive reports below... ««« SOME GADGET GEEK »»» (talk) 20:23, 14 April 2017 (UTC)

Still down. Should we be putting false-positive reports here in talk in the meantime (I've got this one to report), or waiting until it's back up? Ibadibam (talk) 04:20, 17 April 2017 (UTC)
I've been dockerizing and kubifying all of my more recent personal projects, and it's turned out to be really stable. Perhaps it is time to dockerize CBNG and CB3 and move them off wmflabs and on to my Kubernetes cluster. Both CBNG and CB3 were written back in the day before I was willing to pay for hosting, as I was in college and had no money, and the reliability shows. CB3, at some point, should be updated to fix a bunch of the bugs with it. I'm still a huge proponent of the algorithm that CB3 uses to archive threads (based on history, and backlinks), but the code and the execution of that algorithm is lacking, and needs some major cleanup. CBNG's code, however, is still quite clean; it's just the supporting services (report/review interfaces, infrastructure surrounding it) that seem to be less reliable. @DamianZaremba: we should discuss this in the next few weeks. -- Cobi(t|c|b) 05:52, 17 April 2017 (UTC)

False positive

I couldn't find the link, so I wanted to tell you that your revert to Burnin' It Down was a false positive, because I was removing a redlink. --2600:8805:2409:AB00:48C7:1E4:D06C:4085 (talk) 21:16, 8 April 2017 (UTC)

For the record, the link to the diff is here and the MySQL ID is 3000112. —k6ka 🍁 (Talk · Contributions) 01:16, 9 April 2017 (UTC)

  Done Reported using the new link given below. RoCo(talk) 05:53, 21 April 2017 (UTC)

I believe it should go to http://tools.wmflabs.org/cluebotng/ not http://tools.wmflabs.org/cluebot/ . 76.22.118.146 (talk) 22:51, 20 April 2017 (UTC)

That is correct. So the website is not down, it's just the wrong URL. Some folk must have changed CBNG's warning template. If I figure this out I'll bring it back to normal. ««« SOME GADGET GEEK »»» (talk) 01:49, 21 April 2017 (UTC)
I temporarily changed the link of "report false positives" to the above url. If the old url is fixed, it can be changed back. RoCo(talk) 05:56, 21 April 2017 (UTC)

If done by anyone else, I just say that this edit by ClueBot was vandalism, but apparently I'm supposed to call it a "false positive". The link for reporting false positives is not working.

I would not know that ClueBot exists if I hadn't seen it doing things like this, and every edit by ClueBot I've ever encountered that way was worse than worthless. ClueBot should leave mathematical notation alone. Michael Hardy (talk) 00:56, 22 April 2017 (UTC)

  Done I changed the link as per section above, and reported the false positive. RoCo(talk) 05:40, 22 April 2017 (UTC)

What if ClueBot put in the edit summaries its calculated likelihood of the reverted edit's being vandalism?

St. claires fire (talk) 01:19, 24 April 2017 (UTC)

On frwiki, Special:Contributions/Salebot does this already, although it uses a different method of scoring. Furthermore, in the messages it use to warn users, it also identifies what it detected that led to its revert, for example: "Too many vowels in a row; Long passage without vowels; Passage without vowels; 30 characters without blanks; Passage without vowels; Same letter 5 times in succession; Same letter 10 times in a row"... you get the idea. That would be more user-friendly compared to listing just the ANN score in the warning. ««« SOME GADGET GEEK »»» (talk) 03:21, 24 April 2017 (UTC)
It might also give vandals hints on how to avoid triggering the bot. RivertorchFIREWATER 04:38, 24 April 2017 (UTC)
ClueBot vandalism detection is based on machine learning, not on a set of fixed rules; thus even if desired it could not explain its reasons de cette façon: Noyster (talk), 09:03, 24 April 2017 (UTC)
Simply not necessary. The bot puts the ANN score for the reverted edit on the reverted user's user talk page already. This number, ranging from a scale of 0 to 1, shows how likely the bot thinks the reverted edit is vandalism where 1 is when the bot is absolutely sure the edit was vandalism. —k6ka 🍁 (Talk · Contributions) 10:30, 24 April 2017 (UTC)

IRC feed still down

@DamianZaremba: @Cobi: Apologies I haven't followed these postings more closely, but since it seems like you've dealt with some of the issues recently, I thought I'd break this out as a new thread.

The CBNG IRC feed is still down. I know you noted above this was fixed a while back, but looking back, my twice-daily STiki reports would indicate the uptime didn't last too long.

Let me know if there is anything I can do to help. West.andrew.g (talk) 17:42, 24 April 2017 (UTC)

A barnstar for you!

  The Tireless Contributor Barnstar
Yeah I finally beat you to One! Great work all the time- I Am Chaos I Am Chaos 14:27, 25 April 2017 (UTC)

Improper archival of heading, part 3

Further to my four previous posts on this matter (last one at User talk:ClueBot Commons/Archives/2017/March#Improper archival of heading, part 2), none of which were answered (which is against WP:BOTCOMM), I find that ClueBot III (talk · contribs) continues to make these bad edits, and a further problem has surfaced: the bot has archived a thread less than an hour after it was placed on the page. @Cobi: Are these issues going to be addressed? I await your reply: shall I block the bot pending a resolution? --Redrose64 🌹 (talk) 18:52, 24 April 2017 (UTC)

It archived it because it was marked as Done. It is behaving as expected. The headerlevel parameter designates the level of header that is expected for a new discussion, and the bot treats all wiki markup between the first header of the specified level to the next header of the specified level as one discussion. Discussions should all be of the same level header and mixed-level headers denoting different discussions is not supported at this time. -- Cobi(t|c|b) 19:27, 24 April 2017 (UTC)
While not supported, I have added a workaround that should help on the page. -- Cobi(t|c|b) 19:46, 24 April 2017 (UTC)
It didn't work properly, taking the comment opening marker over to the archive. So it looks like I must continue to examine every archiving edit and clean up two pages as necessary. --Redrose64 🌹 (talk) 12:36, 25 April 2017 (UTC)
Also with this one. --Redrose64 🌹 (talk) 20:33, 27 April 2017 (UTC)
At some point when I have more time, I'll add the features which you desire. I thought it could be a quick fix by adding something to the page, but ClueBot III was not designed to handle multiple categories of discussion on the same page. Either the page needs to be reformatted (perhaps into transcluded sub-pages? Those work well with CB3), or you need to figure out a different archive strategy. -- Cobi(t|c|b) 23:32, 27 April 2017 (UTC)