User talk:ClueBot Commons/Archives/2016/September


A barnstar for you!

  The Anti-Vandalism Barnstar
Thank you (ClueBot NG) VarunFEB2003 14:50, 1 September 2016 (UTC)

Question regarding vandalism monitoring

Hello Cluebot - i was monitoring recent pages tagged with possible vandalism to see if they are, actually, vandalism cases. I reverted one case I found and on another one I found I saw that Cluebot had already reverted. But it still showed in the recent pages list. My question (curiosity) is that how will the Cluebot know that a page has already been reverted by a Human. I am very new so I am not sure how a lot of this works, so pardon me if this turned out to be a stupid question. Amiwikieditor (talk) 19:25, 1 September 2016 (UTC)

ClueBot NG detects vandalism by comparing the old page to the new page for every single change to articles. It then generates a ton of statistics including statistics about the different pages, statistics about the user that made the edit, and word and letter frequencies and patterns in both the old and new versions of the page, as well as occurrences of certain symbols and sentence structures. It then feeds all of this information into an AI computer system (like the ones used to automatically filter spam e-mail and the ones used to predict what other items you may be interested in on Amazon based on what you've previously looked at, and the ones used to serve targeted advertisements online) that generates a number, where high numbers mean highly likely to be vandalism, and low numbers mean highly unlikely. If the number is high enough, the robot will edit the article and restore it to the last version before that user edited the page. If the robot runs into an edit conflict, it will abort and leave that to humans. -- Cobi(t|c|b) 20:13, 1 September 2016 (UTC)
Thank you for the explanation. Amiwikieditor (talk) 20:19, 1 September 2016 (UTC)

Adding a trigger phrase?

There has been some ongoing vandalism/promotional addition of a scam tech support phone number to a number of technology-related articles over the past few weeks. It comes from various IP addresses and the perpetrators are taking steps to mix things up enough that something is at risk of slipping through. Is there a way to cause a triggered revert if someone tries to add that phone number or a version of that number? The actual text is being varied and the IP addresses are varied as well. Here is a small excerpt of diffs of the behavior:

Thanks. Neil916 (Talk) 14:47, 1 September 2016 (UTC)

ClueBot NG doesn't operate on trigger phrases, but instead with machine learning. Perhaps try the edit filter? -- Cobi(t|c|b) 20:00, 1 September 2016 (UTC)
I created a simple test filter at Special:AbuseFilter/2. I won't be around much for a while though, I'll try to get Samtar to keep an eye on it. BethNaught (talk) 20:16, 1 September 2016 (UTC)
One eye as requested BethNaught -- samtar talk or stalk 20:18, 1 September 2016 (UTC)
Live at Special:AbuseFilter/793 -- samtar talk or stalk 14:09, 2 September 2016 (UTC)

ClueBot NG down?

Hello all,

The bot appears to be currently down, though it could be fixed by the time people may read this... I believe last time, Rich Smith was able to restart the bot on labs, which fixed it. Though, he appears to be inactive at the time. Anyone else seeing this message and/or patrols this talkpage, could you please take a look at what is going on here...? Thanks! 2607:FB90:A496:F2F:0:4A:4629:8F01 (talk) 16:31, 2 September 2016 (UTC)

I'm always active, just don't always edit :) I'll get the bot restarted - RichT|C|E-Mail 18:34, 2 September 2016 (UTC)
And it is now fixed. —k6ka 🍁 (Talk · Contributions) 18:50, 2 September 2016 (UTC)
...or not. @Rich Smith: Looks like it's down again. —k6ka 🍁 (Talk · Contributions) 00:33, 4 September 2016 (UTC)
Rich Smith, it still looks down. Enterprisey (talk!) 20:24, 4 September 2016 (UTC)
@Enterprisey: And back up again (or is it?). —k6ka 🍁 (Talk · Contributions) 21:59, 4 September 2016 (UTC)
It most certainly is running again. It just reverted a vandalism post in Ancient Egyptian literature. Thanks, bot! I know you're just a robot, but your actions have done so much good. You (and your makers) deserve a little praise. Pericles of AthensTalk 18:50, 5 September 2016 (UTC)

Ethiopian War

clue bot is being used to change history towards a deranged idea of what really happend in the second ethiopian war someone is trying to change it so it looks like italy colonized ethiopia. the ton is wrong and has been flagged since 1997 for bias and i have changed it after many years of research. Still their are people trying to muzzel the truth. — Preceding unsigned comment added by Noahyayeh2011 (talkcontribs) 00:26, 10 September 2016 (UTC)

Hi Noahyayeh. I can guarantee you ClueBot NG is not trying to "muzzel the truth". ClueBot NG is not human, it is a robot and there has been something in your edit that has flagged up as potential vandalism. In all fairness I'd have reverted your edit too because it sounded like you were adding your own opinion to the article and not facts.
As for your request for ClueBot NG never to post on your talk page again, if it has to revert you again I'm afraid a message will be automatically generated. 5 albert square (talk) 02:39, 10 September 2016 (UTC)

Question

Where can one contribute reviewing edits in order to make this bot more accurate? --.Snoyes (talk) 10:12, 12 September 2016 (UTC)

@.Snoyes:It looks as if you can't, at present. If you look at top of this very page, third box down, the item "Review edits for the dataset" links to where this used to be done, but it's not working now. All we can do is report "false positives" where we think ClueBot has wrongly reverted an edit - there's a working link for that: Noyster (talk), 12:20, 12 September 2016 (UTC)
I've just tried to report a false positive (2762019). I've just failed half-a-dozen CAPTCHAs, all of which have been too unclear to input the letters with any degree of accuracy. What now? Pinkbeast (talk) 18:44, 16 September 2016 (UTC)

"Archive failed"

Could someone take a look at these three edits? It seems that ClueBot III tried to archive Wikipedia:Administrators' noticeboard/Requests for closure, but hit a problem. The bot first added the discussions to the archive page, but then reverted the edit with an edit summary "(Archive failed)", then removed the discussions from the main discussion page anyway. Mz7 (talk) 00:57, 17 September 2016 (UTC)

Just a heads up about a little bug. Cluebot archived a bunch of threads at Talk:List of lists of lists, including multiple requested move threads. It then changed the link in the current requested move notice (a link to a discussion that was not archived), to point to Archive 1. — Rhododendrites talk \\ 14:26, 19 September 2016 (UTC)

September 2016

Sir, this is IHitIt1st. My edit originally was acceptable, but something happened to it. I tried to change it, and you are giving me a message. Sir, I believe Suge Knight was born in 1966, not April 1965, so why can't I put that even if I have a source? — Preceding unsigned comment added by IHitIt1st (talkcontribs) 22:28, 19 September 2016 (UTC)

@IHitIt1st: Please read the message in the big red box at the very top of this page. ClueBot NG is a robot and is not a human. If the bot has made a mistake, please report it at User:ClueBot NG/FalsePositives. —k6ka 🍁 (Talk · Contributions) 23:40, 19 September 2016 (UTC)

Wrong warning given

Hey, quick question. ClueBot NG just warned someone with its level one vandalism warning when they clearly deserved at least a level two warning. (See it here.) I realize the first two warnings are seven days old, but, to my knowledge, stuff like this affects what people see in things like Huggle (such as the numbers next to the links to the diffs). (I'm still a little new to Huggle, so I'm not sure about that one.) Can you explain why the user got a level one warning when the page they vandalized is the exact same page they vandalized seven days ago? -- Gestrid (talk) 17:30, 27 September 2016 (UTC)

It's an IP address. It's even a school IP address. The person who got the warning 7 days ago is likely not the same person who this warning is for. The bot ignores warnings over 2 days old. This has been in place since the beginning of ClueBot back in 2007 (wow, it's almost been 10 years ...), and no one has seriously raised (or attempted to raise) consensus for changing it. It would be very easy to change if we decided we wanted it at a different threshold. -- Cobi(t|c|b) 17:56, 27 September 2016 (UTC)
If I sounded harsh, I didn't mean to. I do think it would be better to raise that threshold to (possibly) a week, or, if possible, for ClueBot to check the page names for previous recent (within a week or so) warnings to see if it's warning for the same page or not, then warn them with a higher-level warning if it is the same page. What are the odds that two different people from the same school would vandalize the exact same page within a week? -- Gestrid (talk) 18:02, 27 September 2016 (UTC)
Huggle, or at least Huggle 2, had this same implementation in place. Vandalizing the same page only once a week is considered to be wonderful in the eyes of most experienced patrollers. A school IP address can be blocked for periods of time if vandalism is persistent, but there is no need to mess around with the warnings. FWIW an account does not get this "grace period", and should not, because accounts are believed to belong to one user and one user only. —k6ka 🍁 (Talk · Contributions) 10:41, 30 September 2016 (UTC)