I expect this can only be accomplished using data explorer, but I would like to find comments and answers in response to my content which use pronouns to refer to me.

How do I search for specific words only in comments on my questions and answers?

How do I search for specific words only in answers to my questions?

I'd like this to be network wide if possible, but if not searches can be performed by site.

Note that I've edited the title to be more specific. The two questions above are all I'm really asking, I'm not asking for an NLP implementation - I merely want to review all the posts in response to my posts which contain pronouns.

  • you can forget networkwide, if you have many posts, as there are no full text indexes on SEDE so you only have LIKE '%word%'at your disposal and that lands you in table scan territory. For unknown reasons the comments table hold more rows then the Posts table....
    – rene
    Oct 30, 2019 at 14:04
  • 9
    Simply put, you can't. Finding pronouns referring to you requires a context-dependent analysis. Feel free to download the data dump and run it through an NLP system, but you can't do it with SEDE or any other on-site tools. Oct 30, 2019 at 14:05
  • 3
    Why on earth are these two posts of yours getting downvotes? They are explicitly on-topic for the site, and over the past few weeks (as posted) users have been creating SEDE queries for exactly those purposes. These posts aren't too broad, nor off-topic. They are entirely fine.
    – Bart
    Oct 30, 2019 at 17:04
  • 2
    @Bart I'd hazard folks really believe there's some kind of bad faith behind the question, for some reason. Oct 31, 2019 at 3:10
  • 1
    @Bart Personally, I didn't downvote it but considered doing so because it's really outside of the scope of meta to do this reliably. Following the discussion, I've checked my answers for pronouns, and found some that did, all referring to a generic user of the program and not the OP. You need NLP to determine if the pronoun refers to the OP, and pretty complex NLP at that. Half-answers can lead to misleading stats when used.
    – Erik A
    Oct 31, 2019 at 9:55
  • 1
    @Zoethetransgirl While you're correct that the question title is difficult to answer, the specific steps to get what I want, listed in the question body, are achievable and will allow me to process the content myself. I hope this question is re-opened based on the clarification provided in the question body. I'm not asking for an NLP implementation.
    – Pollyanna
    Nov 6, 2019 at 13:38
  • Nevertheless, I've edited the title as it appears to be confusing people as to the scope of this request. Please consider re-opening this question.
    – Pollyanna
    Nov 6, 2019 at 13:41

1 Answer 1


The following query is meant to get you started with something I invite you or others to expand on. I won't take improvement requests:

select c.postid as [Post Link]
     , c.id as [Comment Link]
     , c.userid as [User Link]
from comments c
inner join posts p on p.id = c.postid
where p.owneruserid = ##userid?2915##
and c.userid <> ##userid?2915##
and c.text like '%##word?wrong##%'

when run today this is what a result looks like, for Stack Overflow:

result of the query for userid 2915 and the word: wrong, showing post links, comment links and the user who posted the comment

Keep in mind that SEDE is only updated once a week on Sunday.
Give a big shoutout to Monica Cellio for her awesome tutorial
Hop into the SEDE Chatroom and say "Hi!".

  • 12
    +1 for the shoutout to Monica, yet another example of the awesome work she's poured into this place over the years Oct 30, 2019 at 14:19

You must log in to answer this question.

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