5

The automatic check I would like to learn about looks like as follows:

For every question posted (alternatively: voted to close or pushed to close queue),

  1. Check if asker has linked SO account
  2. If yes, check 3 of their most recent questions at SO
  3. For every of above 3 questions, calculate Levenshtein distance of its text to that of to Programmers question.
  4. Raise system flag for mod attention (or present blocking message) if distance for a particular question is less than some threshold.

What difficulties could be there in getting above check done automatically?

threshold mentioned above can be derived based on a representative set of already identified cross-posts. This set (few hundreds in last half year), in turn, can be retrieved by searching flag database for my flags that are marked helpful and have message containing http://meta.stackexchange.com/tags/cross-posting/info.


  • Side note (as this has been asked twice in comments). I am somewhat skeptical on blocking message approach. In theory, this would be ideal solution but I am concerned that if presented with block, some users will try to trick to break through Levenshtein threshold (for example, by adding dots to question text). This may return us back to having to discover them manually. My considerations are similar to those why question ban / serial voting algorithms are not exposed for "direct user testing"

marked as duplicate by gnat, Bart van Ingen Schenau, user40980, GlenH7, ChrisF Aug 27 '14 at 7:49

This question has been asked before and already has an answer. If those answers do not fully address your question, please ask a new question.

  • To preemptively address possible comments like oh but system flags will increase moderators load - no it won't. Cross-posts are identified and flagged (just as suggested at MSE tag wiki) and moderators handle them anyway, the only difference would be that flags would raise automatically instead of manually. – gnat May 15 '14 at 14:34
  • 1
    Better yet prevent the user from posting the duplicate and tell them to either flag the original question for migration or delete and repost. – ChrisF May 15 '14 at 14:38
  • @ChrisF I am somewhat skeptical on that. In theory, this would be ideal solution but I am concerned that if presented with block, some users will try to trick to break through Levenshtein threshold (for example, by adding dots to question text). This may return us back to having to discover them manually. My considerations are similar to those why question ban / serial voting algorithms are not exposed for "direct user testing". Anyway, I added your option into the question, let developers consider options – gnat May 15 '14 at 14:51
  • 2
    That's a good point. Extending it to a block was my initial reaction to your idea - if there are good reasons why that wouldn't be a good idea then I'm happy to withdraw the suggestion. – ChrisF May 15 '14 at 15:09
  • This would be excellent, but why a flag? I think a message notifying the OP that there's little value in cross-posting would be good enough. Also #2 should be "If yes, check their last 3 open questions at SO". Lastly, why isn't this a feature request? – yannis May 15 '14 at 19:27
  • @YannisRizos see my prior comment :) "I am somewhat skeptical on that. In theory, this would be ideal solution but I am concerned..." – gnat May 15 '14 at 19:31
  • 1
    Oh come on, you can't expect me to read existing comments before I post my own @gnat ;P – yannis May 15 '14 at 19:33
  • 1
    There is a check to see if a duplicate exists on a destination site which rejects a migration if so, but ... I am not promising anything here beyond that I'll look into it. – Tim Post May 16 '14 at 13:54
  • @TimPost thanks! your insight helped me figure a new, better feature request: Perform automatic checks for cross-posting and question ban at SO when post enters close queue, not only at attempt to migrate. I voted to close this as a dupe of new request as it seems to better cover what I want – gnat Aug 3 '14 at 18:33

Browse other questions tagged .