User talk:ClueBot Commons/Archives/2011/February

Latest comment: 13 years ago by Uzma Gamal in topic Test


Pavao Skalic, "Serbo-croatian" grammer

All of this is futile... there's no cooperation exactly, only the autocratic rule of wiki admins and distinguished users! This rules are made to create deception of rule of fair editing, opened to anyone... I will not comply! There's no use of any discussion, wiki dictators have already made their minds... there's no sincere will to discuss changes in order to improve the article! —Preceding unsigned comment added by 89.164.7.154 (talk) 22:31, 1 February 2011 (UTC)

Changes

Thank you for all the warnings about nonconstructive changes I (we) have made, but this IP address belongs to a public school. There are people who would delete an entire article and write "penis" for laughs, so I think for the best interest of Wikipedia and its many readers that this IP address be blocked from making any further edits. —Preceding unsigned comment added by 69.92.95.145 (talk) 07:58, 31 January 2011 (UTC-8)

Edits by Bhoomikpra

Bhoomikpra had added false fact regarding in history section of the article.He had added 30 AS is a major chak of Rawla mandi.It is totally false because 3 AS is not part of Rawla mandi.Other question raised by above user that Indian wrestling and badminton is not played in Rawla Mandi.This user should get information about wrestler family of Ranjhe Khan and wrestling tournaments in Rawla.Badmindon is popular among students in schools. Shemaroo (talk) 10:59, 3 February 2011 (UTC)

Trivial anonymous edits?

I often check my watchlist for anonymous edits without summaries, and find that they are often (but not always) vandalism. If Of these anonymous edits without summaries, I have seen that all of the valid edits change some letter(s) or punctuation. That is to say, from my experience, an anonymous edit without a summary and changing only digits is always classifiable as vandalism. This crops up in tables[1] and article text[2] in ways that will be particularly unnoticeable to all but the most diligent of reviewers checking citations. Can this sort of vandalism be incorporated into ClueBot? -- ke4roh (talk) 14:13, 30 January 2011 (UTC)

The neural network already includes inputs based in input summary statistics, which include length. Crispy1989 (talk) 14:01, 1 February 2011 (UTC)
Then would a different bot be better suited to the exact scenario I described or is there some tweak in order? It really is time to take care of these edits more explicitly. What's the next step? -- ke4roh (talk) 20:39, 2 February 2011 (UTC)
I think I would interpret Crispy's response as indicating you can't expect ClueBot to have anything explicit added to handle this, although the neural network may actually identify some or all of these as vandalism. If you want to have the explicit check done, it would have to be via some other bot. I think it might well be useful if we could convince the ClueBot NG developers to include an additional input to the neural net: "edit has digits but no alphabetics". This input combined with the other NN inputs would probably end up trained to kill most all the vandalism of the kind you brought up. -R. S. Shaw (talk) 01:08, 4 February 2011 (UTC)
Yes, I like that idea. A slightly more general interpretation for the additional input could be the classes of characters included in the edit - alpha, numeric, punctuation, white space. Thanks! -- ke4roh (talk) 02:36, 6 February 2011 (UTC)

Archives

Seeing that archiving by every 25 messages is a no-go zone, can you archive my talk page every 7 days?--The Master of Mayhem (talk) 17:50, 5 February 2011 (UTC)

See User:ClueBot III for how to archive your page. There is instructions and coding there. Any questions feel free to ask.--5 albert square (talk) 18:02, 5 February 2011 (UTC)

What the...

Why did CBNG give this VOA a level 2 warning when he'd received a level 4 warning just a couple of hours before? HJ Mitchell | Penny for your thoughts? 21:12, 5 February 2011 (UTC)

(talk page stalker) I think because the VOA hadn't recieved a level two warning for that month. For some reason, it was skipped... ~ Matthewrbowker Say hi! 21:43, 5 February 2011 (UTC)

What the hell, I did not vandalise the last time, that was correct from statistics found on the internet put it back before I sue you. —Preceding unsigned comment added by 92.235.233.122 (talk) 20:53, 13 February 2011 (UTC)

come fly with me article

Hi I was irritated that you changed my edit on this article. The article currently states that the character Penny is: 'A first class stewardess for Great British Air, with scathing views of passengers in the first class cabin.' Which is contradictory as you can see. I had previously changed this to 'scathing views of passengers in the economy class cabin' which you changed back to as it reads above. Why did you do this when it is clearly incorrect? — Preceding unsigned comment added by RichYPE (talkcontribs) 16:01, 8 February 2011 (UTC)

If it is a false positive please report it so we can train the bot! DamianZaremba (talkcontribs) 17:47, 8 February 2011 (UTC)

Duplicate level two headings

I recently noticed ClueBot duplicating level 2 headings for the existing month. See User talk:216.174.135.158 for example. GcSwRhIc (talk) 18:40, 8 February 2011 (UTC)

Suggestion for improvement

ClueBot should revert a little more smartly so it doesn't revert back to a revision that is probably vandalism. For example see this revision history at Feudalism:

  • 15:17, 8 February 2011 ClueBot NG (talk | contribs) m (23,639 bytes) (Reverting possible vandalism by 204.38.184.88 to version by 204.38.184.77. False positive? Report it. Thanks, ClueBot NG. (269945) (Bot)) (undo)
  • 15:17, 8 February 2011 204.38.184.88 (talk) (23,669 bytes) (→Definition) (undo)
  • 15:15, 8 February 2011 204.38.184.77 (talk) (23,639 bytes) (→Definition) (undo) (Tag: repeating characters)
  • 15:09, 8 February 2011 204.38.184.14 (talk) (23,628 bytes) (→Definition) (undo)

A vandal logged in from the block 204.38.184.* and made three vandal edits in a row. ClueBlot reverted only the last edit because it saw three different IPs, but it should have reverted all three edits. This should be fairly easy to see: if the previous edit to the one being reverted was made by an IP in the same block, it should also be checked for (or considered as) vandalism. Green Cardamom (talk) 19:15, 8 February 2011 (UTC)

While this is a good idea and technically could be implemented I would raise the following points: a) The bot would have to pull down and check the revision it was going to revert to before doing so thus blocking one of the processed and slowing down the revert b) It would be impossible to check every single edited for if it appeared vandalised. Possibly something like check previous edit -> if vandalised repeat else revert to this version could possibly work but Cobi would need to input on this. I believe the bot will bulk revert edits made by the same IP already but as you say if they are in the same block or happens to be vandalised by another user it would be missed. My conclusion would be that while it would make ClueBot NG more effective and less likely to contribute to vandalism serious thought should be put into the extra processing overhead per revert. The idea in the end is to be very fast and efficient. DamianZaremba (talkcontribs) 19:35, 8 February 2011 (UTC)
The bot sees all three IPs as different users, and I don't think it is feasible to change that point of view. In addition, the bot did not see the first two edits of having enough a large enough score for reversion (they scored low 80s, while the threshold is 85) in the first place, so it should not revert them anyway. -- SnoFox(t|c) 20:26, 8 February 2011 (UTC)

Bot has removed shared IP and block history header

At User talk:204.147.20.1‎. I tried to restore the header; when I did, the bot's message disappeared. I've restored the bot message, minus header. Any reason why the bot did this? Or am I missing something? I have to admit, I'm none too clever with codes, syntax and whatnot. Haploidavey (talk) 19:30, 10 February 2011 (UTC)

ClueBot NG must've had one of those shifty bugs I see pop up every month or two... It blanks a page while trying to make its edit. The error you had made when trying to fix the bot's mistake, was you didn't add the bottom of a template. No worries; I'll fix it up. -- SnoFox(t|c) 18:43, 12 February 2011 (UTC)
Ah! OK, I'm still not at all sure what you did, but thanks for doing it. Haploidavey (talk) 19:18, 12 February 2011 (UTC)
I took the easy way out and reverted ClueBot NG's failed edit, and then re-added the warning. -- SnoFox(t|c) 19:19, 12 February 2011 (UTC)

Possible mistake

The page which allows reporting for ClueBot NG appeears not to be working (at least not for me); so I'll just dump the possibly (I do not speak Afrikaans) erronous reverting here: [3]. Njardarlogar (talk) 17:00, 12 February 2011 (UTC)

My edit

How can my edit on the King Mickey page be unconstructive if I just said redirect it to it's main article? I just want to help this wiki. - 12.176.38.188 (talk) 02:30, 6 February 2011 (UTC)

You redirected to Mickey Mouse, which is a more general article. The section provided more information about the Kingdom Hearts character in specific, from where users can go to the main article about Mickey Mouse. Logan Talk Contributions 02:31, 6 February 2011 (UTC)

My edit was a true edit for the article Taking Back the Night Life(album) for the band Liferuiner. The current edit was done by some immature teenager that mocks the original track names. I cant believe that edit made it through and that my corrections to the real and original track names are not allowed. I am Literally looking down at the cd case in my hands and I know without a doubt that my Edit should be allowed.216.201.66.234 (talk) 09:29, 18 February 2011 (UTC)

Thanks

Thanks you, cluebot! Aallasdfa67usgd60 (talk) 00:51, 11 February 2011 (UTC)

Fixed the edit on the Liferuiner album, thankyou216.201.66.234 (talk) 09:35, 18 February 2011 (UTC)

Why Cluebot NG stopped?

Why Cluebot NG stopped reverting vandalism? 46.152.107.234 (talk) 18:17, 16 February 2011 (UTC)

There was an issue with the server - it has now been restarted. DamianZaremba (talkcontribs) 15:58, 18 February 2011 (UTC)

Cluebot

How does cluebot work?

Morgan kirk. —Preceding unsigned comment added by 71.200.246.61 (talk) 00:56, 8 February 2011 (UTC)

ClueBot NG works by parsing out the change feed and running each diff though an ANN which scores it then depending on set thresholds the bot reverts and logs the change. More information can be found on the user page. DamianZaremba (talkcontribs) 17:42, 19 February 2011 (UTC)

Stop removing True edits

Will you stop removing false edits? We want you brainless loser out of wikipedia by sunset. —Preceding unsigned comment added by 175.38.163.112 (talk) 07:33, 14 February 2011 (UTC)

It's a script. --Perseus8235 17:46, 14 February 2011 (UTC)

Removed a True edit on Liferuiner album Taking Back the Night Life.216.201.66.234 (talk) 09:25, 18 February 2011 (UTC)

If you believe that the edit is valid please report the revert as a false positive - that way we can use the data to train the bot. DamianZaremba (talkcontribs) 17:39, 19 February 2011 (UTC)

Adding duplicate headings on talk pages

As seen on User talk:69.127.206.33 I had previous put a heading on there with a warning. Later the bot issued a warning and made another heading that was the same. Just seems kinda redundant and makes the page look a little sloppy. Please look into it. Alex³ (talk) 02:29, 19 February 2011 (UTC)

Hi Alex, from what I can remember that has already been raised and will be looked into.--5 albert square (talk) 13:46, 19 February 2011 (UTC)
Okay thanks, I sorta skimmed through this page and didn't see. Hope it's resolved soon. Alex³ (talk) 14:01, 19 February 2011 (UTC)
That's ok, it's not still on the page but I do remember it being mentioned in the past. I guess Cobi just hasn't had the chance to look into it properly yet :)--5 albert square (talk) 14:10, 19 February 2011 (UTC)

ClueBot's edits

I was blocking this user tonight and this edit somewhat concerned me. You will see that the user was given a level 4 warning by another user, ClueBot has come along 5 minutes later so on the same date and instead of reporting the IP, ClueBot has gone back to a level one warning? I thought I'd mention it because that's quite serious, the Bot should be reporting the vandalism.--5 albert square (talk) 01:07, 20 February 2011 (UTC)

Nearly-on-the-threshold edits

Have you an online log (possibly IRC, or RSS feed) of such edits whose computed vandalism score is very high but a bit below the threshold?

This would be helpful for two reasons. First, such edit statistically is very likely eligible for manual reverting (rollback), so it could point vandal-fighters to their targets. Second, (few) certainly good edits in that list may become a valuable source for the dataset; this may also be applied to edits of white-listed users which are even above the threshold. Incnis Mrsi (talk) 17:23, 19 February 2011 (UTC)

There are two main IRC channels with ClueBot NG feeds on irc.cluenet.org - #wikipedia-van - this is all *reverted* edits with scores & #cluebotng-spam - this is all *un-reverted* edits with scores. Currently I believe Stiki is parsing out the feed and inserting the entries into there queue for users to manually review along with a couple of other feeds. Feel free to catch any output in those channels for your own use - I use the channels to generate out some rough [graphs]; I doubt these are very accurate but should give you a rough idea of what the scores generally range from. DamianZaremba (talkcontribs) 17:30, 19 February 2011 (UTC)
Thanks, I see #cluebotng-spam, there is quite trivial (for me) to implement filters to reduce traffic. Are you sure that such messages do not exceed 510 bytes each in any circumstances? An edit summary may be up to 250 bytes long, page title and user name also may be quite long.
This set of graphs has a mistake in HTML — five (identical) entries of "notvandalised_reasons.hourly.stats.png" in the rightmost column, instead of "notvandalised_reasons.daily.stats.png" for second row from top, and so on.
Incnis Mrsi (talk) 17:59, 19 February 2011 (UTC)
Oops! I commented out the crontab the other day - should be fixed now. As for the edit summary I can't find the exact code for the feed functions but having just tested a max length message on my usertalk page it seems to handle it fine. DamianZaremba (talkcontribs) 18:13, 19 February 2011 (UTC)
As the author of WP:STiki, thought I could provide some quick perspective. I do consume scores from that feed. Damian misspoke slightly above, #cluebotng-spam shows scores for *all* edits -- including those reverted. Also, messages do sometimes exceed the limit (usually when unicode gets involved). I just try-catch messages in my Java processor. It happens exceedingly rarely. Thanks, West.andrew.g (talk) 21:31, 21 February 2011 (UTC)

Duplication?

This edit seems to have created essentially two copies of the discussion? Perhaps due to page size? 76.121.3.85 (talk) 03:55, 23 February 2011 (UTC)

Portal namespace

I'm concerned that neither the new ClueBot nor the old one is looking at the edits to the Portal namespace. I've being doing a recent changes patrol there for a couple of months now, but a bot would surely be more reliable. For some representative diffs have a look at my recent contributions in Portal space. Some of these pages are only one or two clicks away from the Main Page. -- John of Reading (talk) 20:58, 20 February 2011 (UTC)

Indeed, I second this. Many Portal namespace pages have no watchers and uncaught vandalism sometimes sits for years. It's especially hard to track down vandalism when portals are split up into many subpages. Is there a particular reason why the bot can't watch edits to the portal namespace? Does it require a separate IRC feed or..? -- œ 22:51, 22 February 2011 (UTC)
ClueBot NG is not trained to revert edits on other namespaces. Portal, user, and talk pages contain completely different "looking" content to the bot, and requires a new, classified dataset. If an editor is interested in generating a dataset for other namespaces, speak with Crispy here or on IRC. -- SnoFox(t|c) 23:20, 22 February 2011 (UTC)
Would it be possible to run the old Cluebot algorithm in the portal namespace? There aren't enough edits to generate a large dataset, but a check for swearwords would catch the worst of the vandalism. -- John of Reading (talk) 08:00, 23 February 2011 (UTC)
I don't see why there would be any issue with this - the old bot is based on patterns so would be customizable to the content in the portal namespace. It would be up to Cobi but as the code is opensource there is no reason you couldn't run it yourself if he doesn't have time to maintain it etc. DamianZaremba (talkcontribs) 16:38, 23 February 2011 (UTC)

Duplicate Section Headings

See User_talk:216.56.48.115. Going crazy with the headings, confusing other tools in turn.

I feel like this might have been mentioned before, but it seems like this page archives like crazy. Thanks, West.andrew.g (talk) 18:28, 23 February 2011 (UTC)

Just ask for This

Hello. You have reverted one of my edits. Please explain for this ([4]) (Click the link). There are NO relationship between mucine and isparta. This is NOT CORRECT in the Isparta#International Relations section. Just look for these links. You have been warned. And i'll try to undo. Also, there are many have sister cities of   United States in Turkey without unsourced. (Please reply on my talk page). 85.108.245.48 (talk) 18:37, 23 February 2011 (UTC)

If you believe it is a false positive then please report is so we can use it to train the bot. DamianZaremba (talkcontribs) 18:41, 23 February 2011 (UTC)
  Done -- ke4roh (talk) 18:58, 23 February 2011 (UTC)

Sorry

Sorry if you thought my Edit was vandalism —Preceding unsigned comment added by 81.102.22.148 (talk) 18:43, 23 February 2011 (UTC)

Well the edit is classed as vandalism, that's not constructive to the article at all.--5 albert square (talk) 00:27, 24 February 2011 (UTC)

Did you delete File:Sarah was raped.jpg?

if so, why so? Science editor 2 (talk) 08:27, 27 February 2011 (UTC)

Nope the bot only stalks the main namespace. The bot however did revert your edit adding in the file (Which doesn't exist anyway) - if you believe that is a false positive please report it. DamianZaremba (talkcontribs) 08:32, 27 February 2011 (UTC)

no it must have been deleted by someone else before i added it to sarah walkers page Science editor 2 (talk) 08:38, 27 February 2011 (UTC)

Builtins

Should add HAHAHAHA as a known word as in here. Overall, good performance, just repeating chars are sometimes missed. History2007 (talk) 04:38, 26 February 2011 (UTC)

The bot does not work with static rules like matching for "/(HA)+/i". This would be better suited in a simpler bot, such as old ClueBot or a new bot, if Cobi does not wish to run it again. -- SnoFox(t|c) 05:25, 26 February 2011 (UTC)
None of the other bots caught that. This could be handled if a preprocessor looks for repeating chars and then signals that as a new input field, so it would still be a param-based network, but has a newfield/input called repeat char etc. But anyway... History2007 (talk) 09:20, 26 February 2011 (UTC)
True. Maybe you should speak to Crispy about that one. :) -- SnoFox(t|c) 16:28, 26 February 2011 (UTC)
I forget where his talk page is, but given that you obviously know the idea of the suggestion, could you just paste it to where appropriate? My guess is that he handles "known words" via some pre-process that looks up a database/dictionary anyway, so repeat chars etc. can also be done the same way and will be just one more input - a pretty fatal one for most vandal edits that include it. And amazingly many vandals use the same hahaha chars anyway. Thanks. History2007 (talk) 18:16, 26 February 2011 (UTC)
Crispy1989 has told me that the input for "repeating characters" already exists in the neural network. I suppose it'd be weighed in more properly if we ever get the larger database classified and ClueBot NG trained on it... -- SnoFox(t|c) 18:59, 27 February 2011 (UTC)
I had not looked into the datasets you guys use but my guess is that in time you will be forced into segmented Ensemble learning, as datasizes go up. That always, always happens as a learning project succeeds, the success will lead to the search for more data, and the bottlenecks appear exponentially. But there are almost always strategies to work around them. But that is another discussion. History2007 (talk) 21:43, 27 February 2011 (UTC)

Test

In view of the recent edit history and corrections by ClueBot of the Piano article,[5] I think this edit, changing frequently to Freak Went, Lee, was a test to see whether ClueBot would catch it or perhaps not catch the "derp" change. -- Uzma Gamal (talk) 09:13, 28 February 2011 (UTC)