22
$\begingroup$

When a question is migrated from MO to MSE, the system attempts to retag it using MSE tags (taking into account synonyms on the MSE side only; synonyms are a one-way street). If no MSE tag can be found, then either

  • migration fails, if it was user-initiated; or
  • migration goes through with the question tagged , if it was moderator-initiated.

This was discussed on meta.MO, where Willie Wong suggested

every time a migration fails for this reason, flag down Mariano and ask him to create the appropriate synonym at MSE.

I do not think this solves the problem. For one thing, a MO user whose close vote is decisive is likely to be voting from a review queue, and will not be made aware of the outcome of the vote. Other users, should they come across the question, will see it closed as off-topic, without evidence that it was intended for migration. Looking at the MSE tag synonyms, I suspect that Mariano is not getting a lot of requests of this kind.

Another, and bigger concern for me is that sometimes migration goes through due to some inessential tag being in common, while the essential tag is lost. For example, a question tagged with will end up with only on the MSE side. One question with ended up with . Having such a tag as the only tag is worse for the question than being ; it's a one-way ticket into Tumbleweed land.

I went through last month's migrations from MO to MSE, and found that the following tags were lost in transition, although they could easily be mapped to MSE analogues.

ag.algebraic-geometry -> algebraic-geometry
at.algebraic-topology -> algebraic-topology
co.combinatorics -> combinatorics
fa.functional-analysis -> functional-analysis
gr.group-theory -> group-theory
gt.geometric-topology -> geometric-topology
lo.logic -> logic
na.numerical-analysis -> numerical-analysis
oc.optimization-control -> optimization
pr.probability -> probability

eigenvector -> eigenvalues-eigenvectors
eigenvalue -> eigenvalues-eigenvectors
group-rings -> group-ring
limits -> limit
linear-optimization -> linear-programming
linear-pde -> pde
matrix-analysis -> matrices  
recurrences -> recurrence-relations

Should the above be made tag synonyms?

I emphasize that these are actual examples from migrated questions during the last month only. There are other common prefixed MO tags which did not have a question migrated during this time span.


Added: I'd like to quote Noah Snyder's prophetic words posted on July 21, 2010, the second day of MSE private beta.

it might be convenient to have some similarity in tagging between MO and here so that if and when questions can be migrated directly between them we won't have to do lots of retagging

$\endgroup$
  • 3
    $\begingroup$ I wouldn't try to work around the broken migration tag check, I'd just get SE to disable it entirely $\endgroup$ – Mad Scientist Jan 2 '14 at 6:58
  • 1
    $\begingroup$ @MadScientist But that would not solve the second problem: questions arriving with non-descriptive tags like [reference-request] or [surfaces], the essential tags being stripped from them. All the migrations from which I took the above tags did go through; the check did not stop them. $\endgroup$ – Post No Bulls Jan 2 '14 at 7:19
  • $\begingroup$ If anything, [limit] should be [limits] and [group-ring] should be [group-rings] on MSE as well. $\endgroup$ – Asaf Karagila Jan 2 '14 at 7:50
  • 2
    $\begingroup$ @AsafKaragila Then there are [definite-integral], [indefinite-integral], [extension-field]... where is Atwood's giant $\Huge{S}$? $\endgroup$ – Post No Bulls Jan 2 '14 at 8:03
  • 2
    $\begingroup$ Integrals seem better in singular form; but [extension-fields] does sound better. I only know where is Riemann's giant $\huge S$: $$\Huge\int$$ $\endgroup$ – Asaf Karagila Jan 2 '14 at 8:06
  • 4
    $\begingroup$ Migrated questions need some work with the tags in most cases anyway, I'm not convinced that an automatic solution will be sufficient. I'd rather go for something like a review queue for migrated questions where stuff like missing tags would be fixed. $\endgroup$ – Mad Scientist Jan 2 '14 at 9:57
  • 2
    $\begingroup$ @MadScientist MO and MSE are a special case: the sites share the topic and have similar tag systems; some tags differ in spelling only. In all examples I've worked through, tags coming from MO were reasonable. // I agree that tags should often be improved after migration, but this is more likely to happen when a question already has at least one reasonable tag (because then it's seen by users with expertise in the tag area). $\endgroup$ – Post No Bulls Jan 2 '14 at 17:38
  • 1
    $\begingroup$ Why does MO use such bizarre tags? My general opinion is that this problem is on their end. $\endgroup$ – dfeuer Jan 3 '14 at 14:25
  • 2
    $\begingroup$ @dfeuer MO has good reasons to use prefixed tags, and MSE has good reasons to use non-prefixed tags; this was discussed in the 2010 thread to which I linked. Both systems work fine within each site. The lack of compatibility hits the target site (usually MSE), because it is the one that ends up with inadequately tagged migrated question. From the source site's point of view, there is not a problem: the question has been disposed of (migrated away or closed). $\endgroup$ – Post No Bulls Jan 4 '14 at 2:32
  • $\begingroup$ @dfeuer I somehow agree with you (though MO should keep the tags). But please note that all (or at least almost all) of the prefixed tags on MO have the tag without prefix as synonym. The system should at least take synonyms into account, and then the problem would be a lot smaller. That the system does not take synonyms on the source site into account is in my opinion simply an error in implementation/a bug. I reported it on meta.SO a while ago (though perhaps too indirectly); meta.stackexchange.com/questions/210532 $\endgroup$ – quid Jan 4 '14 at 22:29
  • 2
    $\begingroup$ On a more practical note, if we do not want to wait for a bug-fix, which might never happen, the problem can be fixed on the target site (MSE) via synonyms. That is without changing the displayed (actual) tags while the source site (MO) one would need to change the actual tags. That is, whatever one thinks whose problem it is/should be, the target site can fix it a lot more easily. And also, MO did already the analog thing mainly right upon moving to SE network where synonyms exist, albeit not directly motivated by this issue that then was not yet noted (I think). $\endgroup$ – quid Jan 4 '14 at 23:34
  • $\begingroup$ Isn't this discussion moot as migration can occur between math and physics, statistics, chemistry, tex and whatnot (in both directions)? That would seem to require matching categories among all SE sites. $\endgroup$ – Hagen von Eitzen Jan 5 '14 at 18:34
  • $\begingroup$ @HagenvonEitzen MO->MSE is a special case: (a) "MO is by far the largest migrator to MSE, with 185 questions in the past three months" (source); (b) MSE shares the topic with MO, not with other sites; (c) the tagging system of MSE was influenced by MO and is similar to it. Many tags are identical and are already transferred automatically at migration. But some differ in details: pluralization, prefix, etc, -- and are lost during the migration because of this. Hence the proposal. $\endgroup$ – Post No Bulls Jan 5 '14 at 18:48
  • $\begingroup$ @HagenvonEitzen in addition to the argument of quantity between MO and MSE made by Post no Bulls, let me add that at least the "and whatnot" in your comment is certainly false. The number of sites to which users (not mods) can migrate is very limited on every site. I think at most five is the absolute upper limit (including the sites meta), on MO we can migrate to math.SE and stats.SE (and meta) and that's it. You could find out which sites you can migrate to by starting the process of migration where you get an exhaustive list. $\endgroup$ – quid Jan 5 '14 at 19:53
  • 1
    $\begingroup$ It seems that there was no activity on this for some time. I have suggested a synonym for (functional-analysis), so that users can vote on it. We will see what will happen. (I suppose it would be much easier if a moderator did this. In that case the synonyms would not need votes from other users and they would also have no problem with the sufficient score in the tag.) $\endgroup$ – Martin Sleziak Jan 14 '14 at 12:10
12
$\begingroup$

Let's at least implement this for very popular arXiv-style tags (ag.algebraic-geometry and such) — it's useful both for migration and for (admittedly rare) visitors from MO asking questions here.

On the other hand, creating synonyms for numerous small tags (like recurrences used for less than 50 questions on MO) seems (to me) to be an overkill.

$\endgroup$
  • 3
    $\begingroup$ I agree: getting the mathematical area tags right is vastly more important than preserving secondary, low-usage tags named after mathematical objects. $\endgroup$ – Post No Bulls Jan 4 '14 at 2:33
  • 2
    $\begingroup$ All these arXiv tags, or almost all, on MO have as synonym the/a tag that would work on MSE. IMO the system should be changed to take synonyms into account; that it does not is IMO simlply a bug. See meta.stackexchange.com/questions/210532 for my indirect report $\endgroup$ – quid Jan 4 '14 at 22:33
  • 2
    $\begingroup$ @quid Well, I'm not sure how this could have worked: a question has a MO-tag with 3 synonyms that exist on MSE; what should happen after migration? $\endgroup$ – Grigory M Jan 4 '14 at 22:47
  • $\begingroup$ One could just take the most frequently used (on target site) or all until reaching the the limit five (sorted by frequency), or rather take creation date or alphabet or whatever as deciding factor. Plenty of options that do not seem overly complicated to implement. Added: to avoid potential confusion, as said in comment on question here, the problem is only that synonyms on source site are not taken into account (target is fine). So, my point is not that the proposal here would not work, only that it would not by necessary to do this if what I consider a bug was fixed. $\endgroup$ – quid Jan 4 '14 at 23:10
  • 3
    $\begingroup$ @quid While I agree that the reverse lookup of synonyms could be implemented as you described, and it would be an improvement on what we have now, I am not holding my breath waiting for it to happen. Reverse lookup and dealing with multiple synonyms would noticeably increase the complexity to the migration code. It's not fixing a bug; it's adding a new feature. My impression is that SE is in no hurry to increase complexity of their code base without something they see as a compelling reason. Math sites are just not that important in the SE ecosystem. $\endgroup$ – Post No Bulls Jan 5 '14 at 5:06
  • $\begingroup$ @PostNoBulls in my opinion this is a clear implementation error; the check simply does not do what it is supposed to do namely check if a tag present on q exists on target site. I assume the person(s) responsible for writing it simply forgot about synonyms (and that those on target site are taken into account happens by chance). $\endgroup$ – quid Jan 5 '14 at 13:42
  • $\begingroup$ @PostNoBulls FYI you were right the request and bug report got declined. $\endgroup$ – quid Jan 13 '14 at 11:54
  • $\begingroup$ Being the lazy moderator that I am, can I count on the community to propose those synonyms? Approving the synonyms are very quick. But going through the list and typing them in one-by-one will take some time even for me, and I am unlikely to get through the list working by myself any time soon. $\endgroup$ – Willie Wong Jan 16 '14 at 12:33
  • $\begingroup$ @Willie I'm afraid even this requires some superpowers: when I try to propose 'co.combinatorics' as synonym I get error message 'The suggested tag must exist in the system before suggesting it as a synonym!'... $\endgroup$ – Grigory M Jan 16 '14 at 12:50
  • $\begingroup$ @Grigory: it somehow worked for Martin, but he has a few more rep than you do, perhaps it is a privilege for 10K or 20K. (But even so we do now have plenty of high reputation users.) $\endgroup$ – Willie Wong Jan 16 '14 at 13:03
  • 1
    $\begingroup$ As mentioned in my answer to this question, most of the arXiv tags have been implemented. There are some ambiguities however and those I have left alone for now. $\endgroup$ – Willie Wong Jan 20 '14 at 11:40
7
$\begingroup$

I have keyed in the synonym mappings for most of the arXiv tags.

A few slight mismatches:

ca.analysis-and-odes points now to ; ODE questions will need retagging manually after migration from MO.

ra.rings-and-algebras now maps to .

cv.complex-variables goes to .

qa.quantum-algebra becomes .

kt.k-theory-homology maps to .

gm.general-mathematics does not belong here as a useful tag (the subject of this forum is "mathematics"!); following quid's suggestion it is now an intrinsic tag.

If one sees any rejections due to arXiv tag problems for migrations, please ping me by commenting here. It would be surely due to (a) me having made a typo or (b) my overlooking a tag or two when doing the entering.

$\endgroup$
  • 2
    $\begingroup$ Possibly the rings-and-algebras could simply map to abstract-algebra which, while not as precise a fit, would at least always be an appropriate tag for such a question (on the other hand, I am not sure if ring-theory could ever be inappropriate for such a question, as I have not taken a very close look at the sort of questions with the tag). $\endgroup$ – Tobias Kildetoft Jan 20 '14 at 11:55
  • $\begingroup$ Thanks! For ca.analysis-and-ode I'd go with analysis, which seems catch all. It could still be retagged; but what is key is that it does not get rejected and is not nonsensically tagged. for cv.complex-variables, I would say complex-analysis. Things for complex-geometry will often (also) be tagged dg or ag on MO. For ra I agree with above cmnt. For qa, the closest seems quantum-groups, which I would take (though not perfect). For gm you might ask for this to be made an intrinsic tag of the site. $\endgroup$ – quid Jan 20 '14 at 21:57
  • $\begingroup$ @quid: when you said that you agree with above comment, do you mean the portion where Tobias says "abstract-algebra" or the portion where he says "ring-theory" (or, alas, both)? I'll leave these up in the air for a few more days and hope for more views/comments. If no other ideas present themselves I'll do as you and Tobias suggested. $\endgroup$ – Willie Wong Jan 21 '14 at 8:35
  • $\begingroup$ @WillieWong sorry this was indeed not very precise (tried too hard to stay below 600 char). Mainly I meant to endorse abstract-algebra, but yes likely also ring-theory works fine most of the time (so in some sense both). I would go for abstract-algebra as it should always fit somehow; by contrast strictly speaking, say, linear algebra is in ra.rings-and-algebras which then should not be tagged ring-theory. [This is more theoretical but just to get the idea across that in pronciple ra contains things that are not ring-theory] $\endgroup$ – quid Jan 21 '14 at 10:46
  • 1
    $\begingroup$ Thank you! IMO, this essentially resolves the issue. $\endgroup$ – Grigory M Jan 21 '14 at 14:49

You must log in to answer this question.

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