Hi @AlexanderGruber, could you explain a little more about what you mean by "annoying duplicates", and what sort of examples you're looking for (i.e., do you want links to individual posts, or search phrases, SEDE queries. . .?)
user185131
Just linking your comment on CRUDE below, for completeness:
Request for y'all. I need help collecting examples of: crank questions, annoying duplicates (and associated duplicate targets), and PSQ red flag phrases (such as "i have no idea" or "urgent"). To prevent this from derailing the room, please discuss here.
@Brahadeesh What I mean by annoying duplicates is stuff that tends to be asked over and over ad nauseum and can be easily redirected to an existing question
5
The purpose of this is to help me come up with regex to detect these types of questions.
So, I would love multiple copies, since they may contain different phrasing that I wouldn't have thought of (like 0.999 equals one vs 0.999 = 1)
This is also very important for crank questions because I don't run into a lot of those in my tags but I know some people in CURED do frequently.
user185131
@AlexanderGruber Some highly duplicated questions are problems in popular textbooks like Rudin (in analysis) or Hoffman and Kunze (in linear algebra). I gather all the duplicates when I come across them and vote to close them all as duplicates of one canonical question, usually the oldest. Would such things also fall under your radar? Based on the couple of questions you linked above, it seems to me you want to look for duplicates of a canonical "abstract duplicate", which is a different thing.
@AlexanderGruber A few sentences might also be worth adding into room description. (Option to edit room-info should be available to room owners and to users with diamonds.)
Every tag has frequent tab - that could be a candidate where to search for questions which are asked frequently.
room topic changed to REHAB: Help Alex write regex to autoflag questions. Seeking multiple examples of common duplicates, and phrases that appear in crank questions and PSQs (no tags)
Also if you want to check whether some patterns are noticeable among problematic questions, you check review queue in the low quality posts review queue or some other queues.
I think SD already catches titles which has something like "Mostly non-latin title". All caps title would be reasonable as well. (I am not sure whether such things are caught by SD or not.)
Checking the linked questions is a reasonable approximation of duplicates. I'd guess that something could be also done with SEDE.
@skillpatrol Sorry if I posted too much here. Alexander Gruber asked for ways to find duplicates. The queries I posted seem - at least to me - relevant to that question.
@AlexanderGruber If there are some other simply SEDE queries which would help, let me know and I might try. (I do not know that much about SQL, but after using SEDE quite regularly I think I can do at least some basic stuff.)
@skillpatrol ok, no problem
As an experiment I took this query created by rene to check posts which were/weren't in the HNQ list.
However, it does not seem that it is going to be useful - basically, the words which are at the top of the list seems similar to list of words which are common in titles.
We could try the same query, but order it by percentage of closed questions for the given word. (I.e., looking only at percentage among those questions which contain the word.) However, the ordering we get in this way does not seem that useful either: data.stackexchange.com/math/query/1237873/…
I should explicitly say that those queries with list of words are beyond my SQL skills. I have simply took an existing query and replaced the one place.
I see that I have made a typo above. I meant to say that I changed posthistorytypeid = 53 to posthistorytypeid = 10.
Recently published on the arxiv : https://arxiv.org/pdf/2005.00346.pdf
The paper gives a proof of the Collatz conjecture by analyzing Diophantine equations related to a slight generalization of the Collatz map.
The look of the paper is very professional and seems really well written which gives...
I thought that maybe looking at tags which have many deleted questions could give a pointer which tags are more prone to off-topic posts. I created such query in connection with some previous discussion:
One thing is that for some questions it might be quite difficult to find a reasonable regex, even if they are frequent.
For example, questions about finite/infinite sum of $\sum x^n$ or $\sum n x^n$ might appear frequently. But it is not easy to write down some regular expression, since there are many ways how this can be written, variables can be different, etc.
@AlexanderGruber Any question which is closed as a duplicate should be listed as "linked" in the dup-target's side-bar. If there are a significant number of linked questions, then there should be a link under "linked questions" in the sidebar to a page or set of pages with a complete list. Alternately, you can use the following bookmarklet to open in a new tab the linked questions page for the current question:
@MartinSleziak No, SD doesn't catch all-caps titles. There used to be a detection for that, but it was found to be considerably too prone to false positives. However, SD is very specifically only looking for posts that are spam or rude/abusive. Titles that are all-caps, should very likely be an indicator that the question should get some custodial attention, just not necessarily the content-block which SD aims for.
@Makyen do you know if that includes deleted duplicates?
I'm guessing that will require sql
@Xander very good-- i can do something with required+supplementsry keywords there, like making new proof a required keyword and collatz a supplementary one that increases the crank likelihood score, but doesnt activate without new proof.
@MartinSleziak i dont know anything about vixra-- is that a crankbait theorem like collatz?
Certainly, there are also legitimate works on vixra.
@AlexanderGruber The list you get on the site does not include deleted duplicates. If you check the SEDE queries I posted above, there you can get also list of the deleted ones.
However, the table PostsWithDeleted, which is used there, does not contain titles. (You have to click on each of the links individually to see them.)
