10

This support request will require the assistance of a Stack Exchange employee, probably a kindly Community Manager.

As shown at the bottom of this answer, I present some SEDE results on just how many non-deleted questions’ titles contain any of “correct”, “gramma”, or “right”. I used queries like these to produce those charts:

  1. Closed Questions matching CORRECT/GRAMMA/RIGHT
  2. Open Questions matching CORRECT/GRAMMA/RIGHT

Those allow one to produce a chart like this:

               Closed   Open  Cl/Op%
               ------  ------ ------
1. CORRECT       675    1577     43%
2. GRAMMA        202     515     39%
3. RIGHT          91     293     31%
4. Any           867    2165     40%

However, those figures are quite wrong because they do not contain deleted questions matching those strings.

My specific support request is for a kindly Community Manager1 to please run a version of the closed query against their own private version of SEDE which actually includes the deleted data as well, and then tell us how many they found.

To make it easier, it would be enough just to report back the number of DELETED questions matching any of those strings. With just the result of that single modified query run against the employee-only info, we will be able to fill in the Deleted column’s ??? figure here:

               Deleted  Closed   Open  Cl/Op%
               ------- ------  ------ ------
4. Any           ?????    867    2165     40%

This will allow the ELU community to more realistically consider the potential ramifications of adding those as blacklisted keywords.

Perhaps only “correct” and “gramma’ are what we should end up blacklisting, but we really need to understand the deleted questions first before we can really even discuss this further.

Any other guidance you may wish to give us on this would also be much appreciated.


Footnote

  1. If no kindly Community Manager is available, then any other sort will work, too. :)
4
  • CMs were thin on the ground when I raised it earlier. I'll try again at a better time.
    – Andrew Leach Mod
    Mar 16, 2015 at 13:51
  • @AndrewLeach Thanks. I figured it would take a CM to get at the needed data. It might be a bad idea, it might be a good idea, but without evidence it’s not really possible to know.
    – tchrist Mod
    Mar 17, 2015 at 2:06
  • 1
    Shouldn't it be grammar? I don't get the "gramma", I rarely see this spelling error.
    – Mari-Lou A
    Mar 22, 2015 at 8:56
  • @Mari-LouA I mean partial string matching, not complete words. That way we catch all of grammar, grammatic, grammatical, grammatically and such.
    – tchrist Mod
    Mar 24, 2015 at 3:20

1 Answer 1

7

Including deleted questions:

        Deleted Closed Open Closed% Deleted% Cl or Del% 
        ------- ------ ----  ------ -------- ----------
CORRECT     938   1513 1687     47%      29%        51%
GRAMMA      250    341  333     51%      37%        54%
RIGHT       111     17  294     38%      24%        41%
Any        1299   2031 2314     47%      30%        50%

If a title contains more than one of these keywords, it's only counted once because I used this case statement:

  case 
     when Title like '%correct%' then 'CORRECT'
     when Title like '%gramma%'  then 'GRAMMA'
     when Title like '%right%'   then 'RIGHT'
   end

It's also possible for a question to be deleted and open.

Just for kicks, here's the same query for all questions regardless of title:

Deleted Closed Open  Closed% Deleted% Cl or Del%
------- ------ ----- ------- -------- ----------
  14985  23303 42198     36%      23%        41%

That puts ELU in 6th place when it comes to close rate. (For the curious, the top 5 are Programmers, The Workplace, Network Engineering, Christianity, and Patents.) When compared to the median close rate of 15% (TeX—LaTeX and DBA), the grammar questions look pretty bad. But compared to this site's close rate, titles like '%right%' are awfully similar to other questions.

Also to be considered is if forbidding problem titles will be sufficient to prevent problem questions.

6
  • 2
    So what you're saying is that with just a little work we too can have a top five close rate! :) Apr 7, 2015 at 23:35
  • Jon, thanks very much. So it looks like closed+deleted for "CORRECT" is 59%, while for "GRAMMA" it is 64%. Those were the figures I was most curious about. In building my list of "bad 250 questions", I had separately come to the conclusion that "RIGHT" wasn't anything like good enough at distinguishing questions likely to be closed. Oddly, it seems that it is a good indicator of whether it is going to be deleted iff it is closed in the first place.
    – tchrist Mod
    Apr 7, 2015 at 23:39
  • @tchrist: Oh. Because of overlap, that's not exactly right and it's not ---easy--- possible to get from my numbers. So I ran the query again to add a column for the rate questions are closed or deleted (or both). Does that help? Apr 7, 2015 at 23:49
  • 1
    If more than half of questions containing a word in the title are closed or deleted, Something Is Definitely Wrong.
    – Andrew Leach Mod
    Apr 7, 2015 at 23:56
  • Oh, I see. So all questions are 41% likely to be closed or deleted versus 51% for CORRECT and 54% for GRAMMA. My first reaction is that there probably isn’t enough difference there to justify requesting a blacklist. Were PROBLEM questions on SO being closed/deleted at a rate higher than 25-30% more often than the overall question count there?
    – tchrist Mod
    Apr 7, 2015 at 23:57
  • 2
    @tchrist: 22% of SO questions with "problem" in their titles are either closed or deleted. That compares to 23% of all questions. I've always thought that "problem" was a pretty terrible signal and I guess I know that for sure now. (One problem with this analysis: closing and deletion have increased since "problem" was blacklisted.) Apr 8, 2015 at 0:03

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .