If you go to close a question as a duplicate, CV tries its best to "guess" which earlier question it may be duplicate of. Being statisticians, we are leery of the algorithm, and indeed matching on a few key terms rarely actually finds a good candidate.

Is there a way to flag, save, or index a question as canonical or generic so that we can quickly close duplicates and point users to the better questions and answers?

  • 9
    $\begingroup$ In identifying good candidates, I search on tags, order on votes, and then look for one I like, especially if someone I think is very good at explaining wrote an excellent answer (no names here). But I agree that's a heuristic, not an algorithm. The experience of thinking that this must be a duplicate but failing to find a good thread to specify is surely common. Good grief! Yet another question on discarding outliers.... $\endgroup$ – Nick Cox Sep 18 '19 at 19:11
  • 4
    $\begingroup$ An alternative that I've been thinking about is starting a meta thread aimed at creating a list of canonical threads on various topics. $\endgroup$ – mkt Sep 18 '19 at 19:42
  • 3
    $\begingroup$ One technique I use when StackExchange's search doesn't turn up a good candidate is to employ a search engine to perform a site-search of CV. Because this will also search comments for keywords it often turns up additional hits (or changes their order) $\endgroup$ – Glen_b Sep 18 '19 at 23:03
  • 1
    $\begingroup$ @mkt Such a thread could work but I fear it would quickly get out of hand due to its length. It would also be cumbersome to use: in addition to searching our site, one would have to visit that meta thread and search it (somehow), presumably finding and then following additional links. It would be far more convenient to create a mechanism to help the likely duplicate target appear among the first search hits. $\endgroup$ – whuber Sep 19 '19 at 13:24
  • $\begingroup$ @whuber It might, I'm not wedded to the idea. If getting likely duplicate targets to appear among the first search hits is feasible, it would be great. I thought it would require a lot of extra work at the SE network level and would be a non-starter. And given that the search functionality is so poor, I'd be surprised if they could make that happen. $\endgroup$ – mkt Sep 19 '19 at 13:54
  • 1
    $\begingroup$ Hint about better search strategies here: stats.meta.stackexchange.com/questions/5549/… $\endgroup$ – kjetil b halvorsen Sep 22 '19 at 11:29

There isn't really such a mechanism that I'm aware of.

There may be a partial strategy available if the question is well captured by an existing tag, and you can assume the tags are used correctly (ahem). In that case, you can search on the tag by clicking it, and you can sort the threads categorized under that tag by Frequent. That will give you the threads that are most commonly linked by other posts on CV (think of Google's page rank algorithm). One way that a thread is linked by another is that it is the target of which the other thread is a duplicate. It could also have been a "related: [this Q](link)" comment, which is somewhat useful, or a different kind of comment, which is just noise that obscures the signal. Reading through the top returns can help you find a duplicate target. Regarding your specific question here, the more threads that are marked as duplicates of a single target (or the more "related" comments posted to other questions), the better this will work.


At present there is no mechanism, but the way in which you ask this question suggests we might be able to create one. For instance, we could create (say) a tag and, as a community, make sure it is applied to such threads. (Tags do not apply to individual questions or answers.) Thus, a search for keywords that include "[canonical-question]" would reveal only such threads.


I have tentatively created the tag, provided a Wiki excerpt, and applied it to two obvious threads. In doing so it occurred to me that a good use of this tag would be to identify threads that have been "buried" because they have gone unrecognized and therefore are not highly upvoted. (There's little need to tag the most highly voted questions, because they're easy enough to find.)

One can envision the need to establish clear standards for applying the tag so that we don't have to discuss each case here on the Meta site. In particular, it's not difficult to imagine individuals applying this tag to threads only in order to gain attention to their posts. Among the criteria we might want to discuss are:

  • Whether and to what extent to use upvote counts.

  • Whether controversial posts--as evidenced by downvotes or extensive discussion in comments--should be made canonical.

  • Whether to pay attention to anything other than standard indicators of post quality or popularity (voting, bounties, and views).

One possibility is to make threads canonical only after they have been linked as duplicates to at least one question.

  • 4
    $\begingroup$ My experience with trying to maintain the organization & use of tags on CV does not make me optimistic about this. $\endgroup$ – gung - Reinstate Monica Sep 18 '19 at 18:18
  • 1
    $\begingroup$ @gung I'm being cautious here, but it occurs to me that if such a tag is applied by moderators only when questions are closed as duplicates, the curation might be fairly simple to execute. $\endgroup$ – whuber Sep 18 '19 at 21:21
  • $\begingroup$ I think the idea is good, though I worry that people asking about statistical topics that contain the word canonical may outnumber the legitimate uses of the tag. However, if the concern isn't unfounded I expect we can figure out an alternative (perhaps canonical-thread or something similar) $\endgroup$ – Glen_b Sep 18 '19 at 23:07
  • 1
    $\begingroup$ This answer makes me wonder how many other meta-tags already have a sort of 'function' beyond being keywords for content. For the moment self-study and puzzle comes to my mind. Are there any others? ( I remember once a suggestion in a metapost for a tag to be used for certain type of difficult or challenging questions ) $\endgroup$ – Sextus Empiricus Sep 19 '19 at 1:35
  • $\begingroup$ As @gung, I am skeptical that this suggestion, in the long run, will work as intended. It also explicitly goes against the SE policy of avoiding meta tags, but we do sometimes purposefully ignore it, as with [self-study]. $\endgroup$ – amoeba Sep 27 '19 at 14:33
  • 2
    $\begingroup$ @amoeba I appreciate your concern, but could not conceive of any way to evaluate the potential of this approach without performing some kind of experiment. $\endgroup$ – whuber Sep 27 '19 at 14:36
  • 1
    $\begingroup$ +1 to this, I do like experiments. $\endgroup$ – amoeba Sep 27 '19 at 14:41

You must log in to answer this question.

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