Welcome to the edit filter noticeboard |
---|
Recent filter changes (
):Filter 986 — Flags: disabled
Filter 953 — Actions: none; Flags: enabled; Pattern modified
This is the edit filter noticeboard, for coordination and discussion of edit filter use and management. If you wish to request an edit filter, please post at Wikipedia:Edit filter/Requested. If you would like to report a false positive, please post at Wikipedia:Edit filter/False positives. Private filters should not be discussed in detail here; please email an edit filter manager if you have specific concerns or questions about the content of hidden filters. Click here to start a new discussion thread |
Index 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 11 |
Sections older than 10 days may be automatically archived by Lowercase sigmabot III. |
WP:LTA/BKFIP and "famously" removals[edit]
Wikipedia:Long-term abuse/Best known for IP has been using mobile phone IPs to avoid detection and blocks. They are actively removing the word "famously" from Wikipedia. I'm not sure how to track removals of the word, except when they show up in my watchlist. Any suggestions? Examples: Édouard Stern and Deborah Scaling Kiley. -- GreenC 22:15, 3 February 2023 (UTC)
- They may be an LTA, but those edits look like improvements to the article. One could create a detective filter that tracks removals of "famously" from lines and has some check to make sure that things weren't just redirected, but I imagine that there will be a good number of false positives. — Red-tailed hawk (nest) 05:49, 4 February 2023 (UTC)
- Having a quick look at the LTA page, only two of the known non-IP accounts have ever made it past auto-confirmed, so unless there's a bunch of unknown XC accounts in their sock pile you could use that to filter out at least some of the false positives. You can do a filter to track removals of just the word "famously" and/or the few other variants the LTA likes to remove, but if you make it too specific it becomes too easy to avoid, and if you make it generic enough to prevent that then you'll get a lot of false positives. I dunno. Sideswipe9th (talk) 18:53, 4 February 2023 (UTC)
- The LTA page, and SPI archives, are severely underinclusive. BKFIP is quite recognizable once you've dealt with them a few times so accounts are frequently blocked without formal investigation. Further BKFIP usually abandons accounts once they are suspected, so many socks are never blocked at all. Autoconfirmed accounts occur with regularity, and a handful have even reached extended confirmed. Not that this changes any of the analysis below, but that knowledge may prove useful in the future.
- Disclosure: In the past I had defended BKFIP, but now regret having done so. 74.73.224.126 (talk) 21:05, 6 February 2023 (UTC)
- Having a quick look at the LTA page, only two of the known non-IP accounts have ever made it past auto-confirmed, so unless there's a bunch of unknown XC accounts in their sock pile you could use that to filter out at least some of the false positives. You can do a filter to track removals of just the word "famously" and/or the few other variants the LTA likes to remove, but if you make it too specific it becomes too easy to avoid, and if you make it generic enough to prevent that then you'll get a lot of false positives. I dunno. Sideswipe9th (talk) 18:53, 4 February 2023 (UTC)
- Okay, unpopular opinion: this is not worth our time. These edits are, by and large, improvements. The whole shtick here is to make small mostly-positive edits, get reverted per banned means banned and then insult those making the reverts. By engaging with them, whether by reverting or spending time creating an edit filter we are giving them what they want.
- I have a lot of sympathy with the idea that bans apply to any edits, good, bad, or indifferent. I have G5'd whole swathes of pages at SPI and blocked an awful lot of socks on that basis, however here we have a case where the edits themselves are improvements (honestly, the word "famously" in the contexts here serves no purpose), and reverting will allow the user to have their fun. I recommend we leave these sorts of edits alone and just quietly block the accounts or IPs used when we come across them. firefly ( t · c ) 19:14, 4 February 2023 (UTC)
- Mostly agree. Blindly reverting BKFIP's edits will degrade the encyclopedia. That said, the usual pattern is more like (1) BKFIP makes some edit, with a mean-spirited edit summary. (2) User (who has never heard of BKFIP) reverts or warns BKFIP. (3) BKFIP calls user a "troll" and "vandal". (4) User understandably loses their cool, and says something not 100% civil to BKFIP. (5) Matters escalate to ANI. (6) ANI's resident mall cops get in on the action, and join in with BKFIP attacking user. I don't have a solution here. But I think there's value to monitoring BKFIP's edits with a filter. Please, keep an eye on 667 (hist · log). Yes, maybe 90% of that is false positives. But sometimes you'll be able to interrupt a "situation" before it gets out of hand. Personally, sometimes I don't report the sock if they're sticking to mainspace, not edit warring, not hounding anyone, and not being too nasty. I just watchlist their talkpage (I wish we had a "user watch" feature but I know it would be abused...), and report them as soon as they get into a spat with someone. I never click "mass rollback", though I might revert any talk edits without replies. Suffusion of Yellow (talk) 22:39, 4 February 2023 (UTC)
Why is filter 979 getting so many hits?[edit]
- 979 (hist · log)
See this graph. What's going on here? The VisualEditor problem is fixed. When I open a page in VE and type "1" into the citation tool, it switches to the "re-use" panel, as it should. So where are these irrelevant low PMID citations coming from. My first thought was that people were publishing locally saved copies. But it's been two months now. I'd expect the problem to have tapered off, at least. Any other ideas? Suffusion of Yellow (talk) 21:52, 5 February 2023 (UTC)
- @Suffusion of Yellow If you look at the actual diffs of the edits being made now they seem to be a mix of adding metadata to citations added some time ago, creations of pages that look to be old or locally-stored copies, and false positives:
- [1] False positive
- [2] Was converting an existing reference - it seems like the automatic-switching behaviour may not happen if you try to convert a reference that already exists.
- [3] Citation bot adding metadata to an old citation
- Then there are a few edits to a deleted article from a paid editor who presumably had a local (old) copy
- [4] This seems to be copying an article from a page that already exists
- [5] Likewise
- [6] Citation bot again
- [7] Not sure on this one - long term edit warring or perhaps content from another wiki
- [8] Not sure here either but @Egeymi: fixed their edit, so perhaps they can clarify how the error occurred.
- Does that make sense to you or am I missing something? Sam Walton (talk) 22:19, 5 February 2023 (UTC)
- Yeah, thanks, makes sense. I guess I should have looked at more than a few examples. These refs are going to be shuffled around forever unless they're removed. I still can't reproduce the behavior in the second example, though. Suffusion of Yellow (talk) 22:48, 5 February 2023 (UTC)
- Looking at the ten most recent log entries for the filter, 9 seem to be false positives where the edit in question is correctly citing a very old paper with a correspondingly low PMID (1, 2, 3, 4, 5, 6, 7, 8, 9), and one is an editor putting a DOI as the PMID (1). Sideswipe9th (talk) 22:20, 5 February 2023 (UTC)
- Checked entries 11-20, those are all false positives where an editor is correctly citing an old paper with a correspondingly low PMID. Sideswipe9th (talk) 22:23, 5 February 2023 (UTC)
- I wouldn't call most of those FPs. The citations are totally unrelated to the statements they're claiming to support. Suffusion of Yellow (talk) 22:48, 5 February 2023 (UTC)
- It's possible I was looking at this from the wrong perspective? I was checking if the PMID addition was to the correct paper, ie does PMID 12 correspond to a paper with the title "The effect of adrenaline and of alpha- and...", and not whether the citation itself was correct in the context of the content it was supposed to be supporting. I don't think we have the technology yet to design an edit filter that could determine if a given citation is correct for the context its supporting.
- If all we're wanting to do is to track whenever an editor cites something with PMID 1 to 99, in the main, user, or draft namespaces, the filter will do that. But determining whether that edit was user error, a visual editor bug, or an intentional use of the wrong citation, is something we'd need to start asking about. Sideswipe9th (talk) 23:18, 5 February 2023 (UTC)
- A bit more background. The idea was to warn users with something like this. The tiny number of people who actually intended to cite one of those obscure papers could just click past the warning. Except, now that warning message is wrong. VE doesn't do that anymore. If it's just old refs being shuffled around, and bots filling in missing fields, I guess there's nothing to do. But I suspect that at least a few of these are new, and I'd like to know here they're coming from. Suffusion of Yellow (talk) 23:56, 5 February 2023 (UTC)
- I wouldn't call most of those FPs. The citations are totally unrelated to the statements they're claiming to support. Suffusion of Yellow (talk) 22:48, 5 February 2023 (UTC)
- Checked entries 11-20, those are all false positives where an editor is correctly citing an old paper with a correspondingly low PMID. Sideswipe9th (talk) 22:23, 5 February 2023 (UTC)
Set filter 1241 to disallow[edit]
- 1241 (hist · log) ("User talk page disruption", private)
Mandatory notification. Please do not discuss publicly. Suffusion of Yellow (talk) 22:16, 7 February 2023 (UTC)
"possible vandalism" tag capitalization[edit]
The "repeated attempts to save edit" and "Rapid disruption II" filter currently seems to add the "Possible vandalism" tag (with a capital P), while the other filters (11 (hist · log) for example) add the "possible vandalism" tag (with a lowercase P). This results in two possible vandalism tags. I suggest either this be corrected or changed to something like "possible vandalism- throttled". 137a (talk • edits) 16:05, 15 February 2023 (UTC)
- Nice catch, the tags do seem to be duplicated on Special:Tags. I'm minded to do the fix you propose - changing them to the lowercase - but would appreciate more opinions for a sanity check. ProcrastinatingReader (talk) 12:36, 16 February 2023 (UTC)
- @ProcrastinatingReader they should all just be lowercased, if we actually want different tags use a different term for one. — xaosflux Talk 13:40, 16 February 2023 (UTC)
- Thanks both. Both filters updated. ProcrastinatingReader (talk) 15:15, 22 February 2023 (UTC)
- @ProcrastinatingReader they should all just be lowercased, if we actually want different tags use a different term for one. — xaosflux Talk 13:40, 16 February 2023 (UTC)
Filter about only number changes without summaries[edit]
Hello everyone! I'm an admin from SqWiki. For many months now we're disturbed by a vandal which changes dates and other numbers in articles just for malicious purposes. Can someone help us set up a filter that would catch up edits which contain only number changes without providing any summaries? Maybe you already have such a filter I can copy to my community? - Klein Muçi (talk) 13:40, 22 February 2023 (UTC)
- @Klein Muçi can you provide a couple of example diffs? — xaosflux Talk 14:21, 22 February 2023 (UTC)
- 712 (hist · log) is one that jumps to mind. Otherwise this kind of vandalism (WP:SNEAKY) is a tough problem to deal with in the general case, because a lot of number changes are vandalism, but some are corrections to inaccurate data (or effectively reverts of previous vandalism). AFAIK the best defence against this is recent changes patrolling, because it's very difficult for a filter (or even something smarter like an AI vandalism bot) to know if these changes are legitimate or not.
- It could also be justifiable for your community to disallow unregistered users from making these kinds of changes, but that's more of a community/policy decision, as it involves a trade-off which would also block many legitimate edits. Only applying to edits without a summary is reasonable, but if you can circumvent it just by adding a summary then I imagine it's unlikely to be of use for long. Perhaps a more targeted filter for your particular vandal is possible. ProcrastinatingReader (talk) 15:25, 22 February 2023 (UTC)
- ProcrastinatingReader, @Xaosflux, hello and thank you for your expressed interest! It so happens that we have already discussed this matter together a few months ago: Special:Diff/1091973424#mw-diff-ntitle1
It's the same problem persisting. You sayIt could also be justifiable for your community to disallow unregistered users from making these kinds of changes, but that's more of a community/policy decision, as it involves a trade-off which would also block many legitimate edits.
and that is the exact situation in which we are. Being a small wiki, we generally don't allow small number only edits by IPs without summaries even when we don't have a strong reason to suspect they might be vandalisms, the reason for that being that we lack the active volunteers needed to check and approve each of those edits. So that is a trade-off we have already been accepting for more than half a year now. The filter would just make de facto manual work that has been already happening for a long time easier. If you find the courage to travel on the rabbit hole of links I've brought above you'll see a lot of more details for this situation which is a desperate recurring request from me now. You'll also see that some MediaWiki changes we're issued to accommodate such a request but all started discussions were closed without any fruitful results. — Klein Muçi (talk) 17:01, 22 February 2023 (UTC)- @Klein Muçi this ask seems a bit narrower in scope, can you provide a couple of diffs that are still in recent changes that illustrate the current edits? — xaosflux Talk 17:15, 22 February 2023 (UTC)
- Xaosflux, I've currently blocked for the millionth time the IP user which usually does these kinds of edits so I'm not sure if there are any on recent changes left (90% of those edits for us come from that one single user) but maybe I can find some old ones or imitate them artificially in my sandbox? Would any of those methods benefit you? — Klein Muçi (talk) 17:22, 22 February 2023 (UTC)
- @Klein Muçi In last 90 days should be fine, even if already reverted (so long as not deleted). Reason why is so that the AF Examine tool can be used to get the most details out of the edit. — xaosflux Talk 17:25, 22 February 2023 (UTC)
- Xaosflux, yes, I understand. I'm guessing some sandbox artificial edits would also help then, no? That would be easier to achieve than finding past edits. — Klein Muçi (talk) 17:28, 22 February 2023 (UTC)
- @Klein Muçi they could, but this was also to attempt to determine if how you describe the situation above is actually what the situation is; rules are programmatic but memories can be fuzzy :D For example is it ONLY strict number changes such as "156156" --> "156256", or is it also things like "100 003,1415" --> "100 003,1115", "3,14" --> "31,4", and "tetëmbëdhjetë" --> "gjashtë". Is it really null edit summaries, or is it also edit summaries that aren't null, etc. — xaosflux Talk 17:49, 22 February 2023 (UTC)
- Xaosflux, Edit 1, Edit 2, Edit 3.
These are all artificial edits. The first is full page edit with one single page, the second is full page edit with multiple changes and the third is a section edit (is this really "without summary"?).
I fully agree with your logic. Such variations do indeed exist for us but the idea is to hopefully get the code/regex for the general case and further modify it as needed. If you wish to go the extra mile, you might present me with the "general regex" and a variation where space + commas/periods are introduced. I believe that would be more than enough for me to model on further changes. — Klein Muçi (talk) 18:59, 22 February 2023 (UTC)- OK, that matches: 6358179, 6358180, and 6358181. — xaosflux Talk 19:07, 22 February 2023 (UTC)
- @Klein Muçi so for example, notice that third one: it does have an edit summary:
/* Filmografia */
. — xaosflux Talk 19:08, 22 February 2023 (UTC)- Xaosflux, I see. Any way to include such changes that have no manual inputted summary?
If not, let's stick with whole page edits, only number changes without summaries by IPs (even though the artificial examples I brought were made by registered accounts). — Klein Muçi (talk) 01:10, 23 February 2023 (UTC)- Sorta, the "only number changes" is the hard part of this (and a MVP) - so that's the first hurdle, lets see if anyone comes up with any bright ideas for that first. — xaosflux Talk 01:52, 23 February 2023 (UTC)
- I commented on phab:T220764 that having better diff info to AF may solve this. — xaosflux Talk 15:51, 23 February 2023 (UTC)
- Sorta, the "only number changes" is the hard part of this (and a MVP) - so that's the first hurdle, lets see if anyone comes up with any bright ideas for that first. — xaosflux Talk 01:52, 23 February 2023 (UTC)
- Xaosflux, I see. Any way to include such changes that have no manual inputted summary?
- Xaosflux, Edit 1, Edit 2, Edit 3.
- @Klein Muçi they could, but this was also to attempt to determine if how you describe the situation above is actually what the situation is; rules are programmatic but memories can be fuzzy :D For example is it ONLY strict number changes such as "156156" --> "156256", or is it also things like "100 003,1415" --> "100 003,1115", "3,14" --> "31,4", and "tetëmbëdhjetë" --> "gjashtë". Is it really null edit summaries, or is it also edit summaries that aren't null, etc. — xaosflux Talk 17:49, 22 February 2023 (UTC)
- Xaosflux, yes, I understand. I'm guessing some sandbox artificial edits would also help then, no? That would be easier to achieve than finding past edits. — Klein Muçi (talk) 17:28, 22 February 2023 (UTC)
- @Klein Muçi In last 90 days should be fine, even if already reverted (so long as not deleted). Reason why is so that the AF Examine tool can be used to get the most details out of the edit. — xaosflux Talk 17:25, 22 February 2023 (UTC)
- Xaosflux, I've currently blocked for the millionth time the IP user which usually does these kinds of edits so I'm not sure if there are any on recent changes left (90% of those edits for us come from that one single user) but maybe I can find some old ones or imitate them artificially in my sandbox? Would any of those methods benefit you? — Klein Muçi (talk) 17:22, 22 February 2023 (UTC)
- @Klein Muçi this ask seems a bit narrower in scope, can you provide a couple of diffs that are still in recent changes that illustrate the current edits? — xaosflux Talk 17:15, 22 February 2023 (UTC)
- ProcrastinatingReader, @Xaosflux, hello and thank you for your expressed interest! It so happens that we have already discussed this matter together a few months ago: Special:Diff/1091973424#mw-diff-ntitle1