Not signed in (Sign In)

Vanilla 1.1.9 is a product of Lussumo. More Information: Documentation, Community Support.

    • CommentAuthorBen Webster
    • CommentTimeNov 8th 2009 edited

    While MO has a very nice tagging system, many people who I otherwise think of as intelligent and congenial users are clearly not putting any thought into choosing tags; for example, not to pick on Pete, but he posted a very nice question which he then proceeded to tag "topology, algebra" as though that would help anyone sorting through questions, rather than something actually useful, like "abelian groups" or "geometric group theory."

    How do we encourage people to use better and more specific tags?

    For the specific Pete, this was his first question, and the tag system, as many things are, requires some experience.

    Here's my perspective about the tags: I think they should be imagined as allowing different "sub-sites" within the site. That is, somebody might decide to follow mostly "algebraic-geometry" tag and set up his home page to that tag. Or, an RSS reader can be asked to download "topology or algebraic-topology" expression-tag.

    Interestingly, while you added an excellent tag "algebraic-topology", I'm not sure the other 4 tags you added are that relevant either: "abelian-groups fundamental-groups free-groups geometric-group-theory" all are created for this question, which kind of defeats their purpose: there still aren't any people monitoring "free-groups".

    Retagging is one thing that we should encourage people to do once they get the right to do so (I do retag a lot). This operation requires just 500 rep, so perhaps we should just make it more well-known to people who have that much reputation that they can retag (responsibly).

    Here's a Tagging FAQ from meta.stackoverflow.

    By the way, I remember I heard once that Community User automatically retags some questions. Do we have it on MO?

    • CommentAuthorBen Webster
    • CommentTimeNov 8th 2009 edited

    Well, somebody has to create tags. Nobody's monitoring geometric-group-theory or free-groups yet because people haven't bothered to use it. We have a large number of talented geometric group theorists running around the site, so hopefully they'll start using the tag.

    I hadn't noticed that it was Pete's first question (he has done a lot of answers and has been on the site for 2 weeks), but that's certainly not the only example. Another good one I just retagged is Qiaochu Yuan's which was about interpreting super-vector spaces in terms of Hopf algebras, and didn't use the quite well-established tag for "hopf-algebras." Qiaochu is about as active and engaged as user as we have, so it worries me if he's not using tags like that.


    By the way, I retagged that question down to two existing tags. If we differ on that, I suggest to discuss specific tags here. I also hope other people will join this conversation.

    It's a good point that tags are underappreciated. My guess would be our users are still able to read all questions, so people don't uses the list of questions under given tag that much. I don't use it a lot either, though I use tags for highlighting ofinteresting and skipped questions in the question list.

    • CommentAuthorrwbarton
    • CommentTimeNov 8th 2009

    I think Ilya is right that in the long run the usefulness of tags will increase, and people will put more thought into them, as the site gains more questions and more users. I do retag questions on occasion, and I'll try to keep an eye out for questions with poor tags in the future.

    One question I have is, if a question does not directly mention a concept, but the (or an) answer does, does it make sense to later tag the question with that concept? e.g. r-matrix here, or maybe spectral-sequences here.

    I don't know if this is quite the right place for this, but I notice that we now have tags "supersymmetry" and "super-linear-algebra". The latter is very specific, and the former seems to mean something specific to physicists which isn't quite what the questions without that tag are about. Rather than create "super-linear-algebra", "super-algebra", "super-manifolds", etc., maybe just go with a single tag "super-stuff"?


    I'd quite like to keep the explosion of tags a little more contained. At the moment, I'm using tags as a first-line way of figuring out which questions I might be interested in: my favourites are highlighted in some garish colour. But because everyone keeps making new tags, I have to keep updating my list rendering it practically useless.

    Perhaps the tags should be hierarchical. So if someone asks a specific question on, say, spectral sequences then it gets tagged "spectral-sequences" but also tagged "homotopy theory" and "topology". Then by following "topology" I get all questions of a topological nature and don't have to keep track of sub-fields and the like.

    Greg Kuperberg complained to me about the tagging system yesterday. Two proposals I liked were

    - define a "retract" to the arxiv subjects
    - require that every post has an arxiv subject.

    Both would probably require software support, although I can imagine hacking the second one through JavaScript.

    He also said, and I agree, that it was a big mistake to not introduce the arxiv tags before inviting users. Maybe it's just too late now. I might be possible to arrange a "re-tagging day" to coincide with a banner asking people to tag with arxiv tags.

    We could consider trying to write a "tag policy enforcing robot" that downvoted and commented on poorly tagged questions. Pity about the Turing test.

    Scott, I agree questions should have at least one "area" tag, like algebraic-geometry, number-theory, quantum-groups and so on.

    I think there are three things to be done:

    • tag system could be clarified, and perhaps a list of top-level tags (extending the above) inserted into FAQ
    • people can write a canonical retag FAQ, similar to the one at SO
    • users with >500 rep should be reminded they can retag

    Still, I would not expect that people who ask questions can tag them properly -- often knowing the correct classification would be a big part of an answer! Rather, I would enlist help from 50+ users who currently have the ability to retag (it's a good topic for a separate FAQ for them)

    I just went through the first 50 unanswered questions and added any tags that I thought were missing. If just a few people did things like this occasionally, it would probably help a lot. Maybe we should use the top banner (the place that says "you have 3 new responses") to suggest useful tasks for our frequent users.

    I like the idea of encouraging 500+ rep users to retag liberally much better than trying to "solve the problem in software". Of the 54 users capable of retagging, 27 of them have tried it. I'll look in to whether it's possible to show a different tip (something about retagging) to 500+ rep users than to everybody else. I don't have control over the banners that appear when you have new responses, but I do like the idea of using them to encourage high rep users to perform useful tasks. One problem is deciding when they should appear; I think it would get annoying if they appear more than once a week. For now, we should encourage people in comments to retag and lead by example. If somebody leaves a comment on a question saying that it should be closed or converted to wiki, I've been leaving a comment explaining that in the future they should flag such a post for moderator attention. But I'm not really sure how or when to leave comments to encourage people to retag.

    @rwbarton: I think it makes sense to retag something if one of the top answers uses a concept, even if the original question doesn't mention it.

    @Scott and @Ilya: I've modified the ask page to include the following text under the tags box, but I don't think it's very good. How should it be modified?

    Please use at least one of the arXiv subject area tags: algebraic-geometry algebraic-topology analysis category-theory combinatorics commutative-algebra complex-variables control-theory differential-geometry dynamical-systems functional-analysis topology geometric-topology group-theory information-theory k-theory homological-algebra lie-algebras lie-groups linear-algebra logic mathematical-physics metric-geometry number-theory numerical-analysis odes operator-algebras optimization pdes probability qft quantum-algebra representation-theory ring-theory spectral-theory statistics subfactors symplectic-geometry tqft derived-category geometry

    If people have suggestions for tags that should be merged or renamed, please post on the Tags to Merge thread.

    I like Ilya's idea of making a retagging FAQ. I, for example, don't feel entirely comfortable with the tagging system; it feels a little too chaotic to me, and I wish there were some rules (or at least heuristics) I could easily apply. The meta.SO tagging FAQ Ilya linked to is pretty good. The problem is that I'm not sure where to put a FAQ. I feel like there should be a way to use meta.MO to run a wiki-like FAQ. Poking around the settings, I see that it's possible to create roles that only have access to certain categories. It would be nice to make a FAQ category and give forum members the ability to edit posts within that category, but I don't see how to do that. If somebody knows how to do this, please let me know.


    Along the lines of Greg Kuperberg's suggestion, Andrew Stacy says on another thread:

    I guess ideally one would be able to define a hierarchy of tags so that [topology] "owned", say, [homotopy theory] and searching by a high level tag brought up all the lower ones as well. This would make sense outside mathematics - has anything like it been tried on the other SE sites?

    I'm pretty sure it hasn't been tried on any SE site (there's no support for it) or on the main S[OFU] sites. The problem I have with this approach is that it doesn't seem sustainable. When somebody creates a new tag, how should it be decided which tag "owns" it or which arXiv tag it "retracts onto"? I feel like tagging is fundamentally something for humans to do; it's how we humans do our part to help out with things computers are bad at. Given that humans have to be involved in tagging, I think it should be kept as simple and non-hierarchical as possible. If the tagging system becomes something more complicated than what it is now, even fewer people will put in the 15 seconds it takes to use tags properly or retag a question because they don't understand the system. Low barrier to entry is absolutely essential for crowd-sourcing.

    I don't mean to say that we shouldn't aspire to have a better tagging system, but if we're going to do something fancy, it had better be for a really good reason, not just because it sounds good. It's easy to get carried away with a beautiful architectural idea of how vast amounts of data should be organized, forgetting that you have to continuously maintain it and incorporate new data.


    I've no idea who this "Andrew Stacy" person is, but he sounds like someone worth listening to.

    Seriously, a hierarchy of tags could easily be implemented. Whenever a user puts in a tag that hasn't been seen before, they are asked for a "parent" tag. Since they created the new tag, they have a vested interest in making sure that it fits in well with the current tagging system. So there'd be no need for an automated system.

    However, the hierarchy is pie-in-the-sky as I guess that it would involve deeper coding than you have access to. The serious point is that when looking for a good implementation, it can be useful to look from both directions, even if the "ideal" (whatever that is) is unachievable.

    A system that would be implementable would be to insist on at least one "top level" tag for each question. "Top level" could correspond, roughly, to the arXiv classification, I guess. I know nothing about javascript, but presumably it would be easy to check if such a tag is selected and prompt for one if not.

    On a slightly different note, I have to confess that I don't really understand what the tagging system is for. I mean, I have a vague idea but that's not really enough for me to select the right tag. As a questioner, the point of the tags is to make sure that my question is seen by someone likely to know the answer. Of course, the strategy then is to select all tags! But I acknowledge that if everyone did that, the system would become useless. So I need to select more carefully, but to do that I need to know how people are using the tag system from the other end. And that I don't know. At the moment, the number of questions is low enough that I'm using the tags merely to highlight potentially interesting questions. Later on, I may start seriously filtering by tags and ignoring anything, say, with [algebraic-geometry] on it. But I have no idea whether or not anyone else is using them this way, or maybe in a different way. Some people might be fairly restrictive: after all, my true area of expertise isn't all that large so if I restricted solely to that then I'd only check one or two tags. But my wider area of "knowing a bit more than the average mathematician" is much broader so I probably ought to be checking in this category. What are other people doing?


    One thing I sometimes use the tags for is narrowing down a search for a question I know I've seen before. For example, if I know the question I'm looking for had the tags [commutative-algebra] and [examples], then I'll visit the URL examples

    (Aside: you can also use the search box to do this. For example, a search for "[commutative-algebra] [algebraic-geometry] stack" will produce a list of questions that use those two tags and contain the word "stack". Annoyingly, if your search only contains tags, then you get a broken link. If you replace all the +'s in the broken URL with spaces, it works as expected.)

    If I retag something, it's likely to be because I want to be able to find it this way later.


    Here's the whole text under the box:

    Combine multiple words into single-words, space to separate up to 5 tags (algebraic-geometry logic numerical-analysis
    Please use at least one of the arXiv subject area tags:
    algebraic-geometry algebraic-topology analysis category-theory combinatorics commutative-algebra complex-variables control-theory differential-geometry dynamical-systems functional-analysis topology geometric-topology group-theory information-theory k-theory homological-algebra lie-algebras lie-groups linear-algebra logic mathematical-physics metric-geometry number-theory numerical-analysis odes operator-algebras optimization pdes probability qft quantum-algebra representation-theory ring-theory spectral-theory statistics subfactors symplectic-geometry tqft derived-category geometry)

    Which I would change to

    Combine multiple words into single-words, space to separate up to 5 tags (algebraic-geometry moduli-space k3)

    Please try use at least one of the tags describing arXiv subject area: algebraic-topology for math.AT, dynamical-systems for math.DS, logic for math.LO, and so on, or a similar widespread tag, such as derived-category, tqft or soft-question. Consult Tagging FAQ for more info.


    Since we don't have Tagging FAQ yet, I would link the last phrase to the SO tagging faq from above.


    I've changed it to this:

    Combine multiple words into single-words, space to separate up to 5 tags (algebraic-geometry logic numerical-analysis)

    Please try use at least one of the tags describing arXiv subject area: algebraic-topology for math.AT, dynamical-systems for math.DS, logic for math.LO, and so on, or a similar widespread tag, such as derived-category, tqft or soft-question. For more on how to tag you question effectively, consult this Tagging Guide (via meta.SO)

    Annoyingly, I only have control of the text between the first open paren and the last close paren.

    btw -- when I've been talking about "arxiv subject areas", I think these should be strings like math.NT, math.GT, etc, just like Terry Tao uses. I know this breaks with current practice, and perhaps is a bit opaque to people who aren't familiar with the arxiv, but I'd argue that being unfamiliar with the arxiv correlates pretty well with not being ready for MO.

    @Scott: the thing I don't like about that is that only two of the characters communicate any information, and I'm never completely sure that I have the right two characters. Perhaps we could compromise and rename [number-theory] to [NT-number-theory], and similarly rename the other arXiv tags.


    I'd argue that being unfamiliar with the arxiv correlates pretty well with not being ready for MO

    I'm not familiar with arXiv in this sense, since I only remember a few easy subject areas, like AG, NT, LO, RT (this already has a potential a confusion between Representation Theory and Ring Theory). In my opinion there is a clear benefit of allowing people like me to easily recognize all areas by a long name.

    Currently, lots of activity happen under derived-category, homological-algebra and soft-question tags which won't fit arxiv-style but seem to me to be valid top-level tags. Next year, the top tags could be different. There's a gain in not tying our hands to arXiv unless the benefits of such a move are proven.


    I don't really care what the top-level tags are so long as they fulfil the following criteria:

    1. They cover just about every possible topic in a small number of tags. (Insert mandatory joke about compactness here)
    2. They don't change. (Too often)
    3. They are used.

    So far I've answered far more questions than I've asked so I'm much more concerned with tags from an answerer's point of view, in which case I just want to bookmark some tags and then forget about them. So I don't care what they are, so long as they select the right topics. If everyone tags their topology questions with "Teapots" then I'll bookmark "Teapots" and be happy.


    The more I think about it, the more I like the idea of renaming the arxiv area tags to have the two capital letters at the front (like [RT-representation-theory] and [LO-logic]). It doesn't make them hard to understand, and it visually distinguishes them, so when you're looking at the homepage, its easy to spot the questions that don't have a tag starting with two capital letters, so you can easily retag them. If nobody objects to this, I'm probably going to start renaming the tags this afternoon, or tomorrow.


    Extremely minor feature request: would it be possible to have retagging edits labelled as such: so "retagged by X" rather than "edited by X"?


    I have to admit I'd love to know what other questions are going to match the tag "Fields medal", Ben!

    I suppose that if Terry Tao asks a question on MO, he could use it ...


    @Anton, I like the idea of having visually distinguished top-level tags.

    However, right now there are quite a few good top-level non-arXiv tags (e.g. derived-categories or soft-question) and there could be more in the future. What to do with them?

    There are several possible choices of action -- none is terrible from my point of view, but I would say the topic deserves a top-level post and some thought going into it.


    @Andrew on retagging/editing: You should post that feature-request on meta.SE.

    @Ilya: I'd just leave non-arxiv tags alone. As a rule, a question should have an arxiv area tag. Sometimes the question is valid but no arxiv tag fits. That's not a problem, but it shouldn't happen too often. I'm not proposing rigidly enforcing the "must have an arxiv tag" rule, but encouraging it as a social convention and making it easy to retag to maintain that social convention.

    I should point out that renaming a tag is a trivial matter for moderators. We don't have to do anything crazy like manually go and change all the tags. This also makes it easily reversible.


    @Anton, let me explain an example scenario that I would like to see on MO.

    I would imagine it possible that a breakthrough in understanding something happens. For the purposes of the hypothetical example, let it be a fundamental lemma, describable by tags "algebraic-geometry geometric-langlands fundamental-lemma". So, several dozens of people post lots of questions tagged to that extent, e.g. 200 questions.

    During that process, each individual question is posted with a knowledge of previous ones. Therefore, at some point somebody is tired of putting all three tags (+somebody following "algebraic-geometry" complains about the proliferation of fundamental-lemma questions) and that person might just decide to post under only "fundamental-lemma", rather then under all tags. Others then shift behavior and start to use the remaining 4 tags to fine-grain their question within a big topic.

    Without any guidance about tagging, the discovery process works, roughly speaking, under the rules of the free market, that is, for the question/answer pair to be successful, an interaction between only one poster and one asker is required (the more people involved, the better, but it's still a voluntary exchange between consenting adults). This is not the optimal social equilibrium: some guidance useful to other people is worth giving, so it makes sense to force/nudge people toward putting universally recognizable tags.

    But too much nudge, I worry, might kill the abovementioned discovery process, where from the everyday work of dozens of regular people suddenly a new behavior -- the radical change -- arises, in this case in the form of a new top-level tag. With too easily recognizable top tags, of course, everyone would continue to put the same AG-algebraic-geometry on the fundamental-lemma until eternity. Why this could be less than optimal? Just look at "high energy physics - theory" archive.

    These comments should be interpreted as just general thoughts, so, sure, I'm fine with retag. Just something to think.


    @Ilya: I think the two of us agree that the system should be flexible enough to allow for organic development. That's why I don't very much like the idea of making it absolutely mandatory to use at least one arxiv tag. But I don't think that renaming [algebraic-geometry] to [AG-algebraic-geometry] will have the effect you're describing since people who haven't read this thread don't think of the tags as being in any kind of a hierarchy. I think the pervasiveness of tags with those two letters in front will just remind people that they should use at least one broad tag.


    @Andrew - at least it was an improvement over "fields"!


    By the way, I'll just note-tags refuse to stay capitalized at the beginning. Anyways, I added the ag- to algebraic geometry to see how it looked. I'll admit, I'm not impressed.


    @Anton: on retagging/editing - I suspected this would be an SE issue, but thought it worth finding out if it's just me or not.

    @Ben: Okay, okay, I surrender ...

    On cases, is it possible to make tags case insensitive? Or shall I put that over at meta.SE?

    (On that note, whilst I have no wish to "game" the meta.SE system, I think it makes sense to spread the load a little of making feature requests and so forth as it will give the meta.SE people an idea of how popular mathoverflow is compared to other sites. However, I'm hesitant at just heading on over and spouting forth about how much better mathoverflow would be if we just had this or that new feature which is why I would tend to bring them up here first to get a sense of whether or not it's just me or not. Not being at Berkeley, I'm not surrounded by hordes of other MOwers to chat with over tea - the one other NTNU person that I know of who lurks here is on sabbatical Somewhere Else!)

    @bwebste, re:capitalization. How about we use math.AG-algebraic-geometry, etc?

    @Scott: does that work? as far as I can tell, tags refuse to be anything other than lower-case.


    I can sort of understand why tags are always in lowercase (to avoid proliferation of [LOOK-AT-ME] and stupid distinctions like [Algebra] vs [algebra]), but I'd really like caps for this purpose, so I made a feature-request at meta.SE.

    Edit: @Andrew: I actually made the feature-request before your comment about distributing feature-requests. I think it's fine for anybody to just post bug reports and feature-requests as they come up, without worrying about distributing the list of askers. Of course, it helps if the request is relevant to other SE sites (or can be made relevant to them), but don't worry about that. If other SE users don't think it's important, they won't vote it up. On that note, perhaps we should reserve some bug reports and feature requests for people who don't have enough reputation to vote on meta.SE.

    Of the people who frequent meta.MO, here are the ones I know of who are on meta.SE: Scott Morrison, Andrew Stacey, Ilya Nikokoshev, David Brown, and myself. I encourage everybody else to head over there and earn a little rep, since meta.SE votes help get features implemented and bugs fixed.


    The helpful message about tags includes the following:

    For more on how to tag you question effectively


    The problem with heading on over to meta.SE is that it seems (at first glance) to be somewhere to record bug reports and vote up others bug reports (and feature requests). There doesn't seem much opportunity to earn rep by answering questions. That's why I suggested distributing the reports a little, as this is our little horde of opportunities to gain a little rep over there, and the only reason I can see for having rep on meta.SE is to vote up other posts for the attention of the guys actually writing the software.

    (Just in case one of them drops by here, I hope this isn't interpreted as me/us trying to "game" the system. If it is, I'll instantly withdraw the suggestion. My intention is purely that I want us to be able to convey how much we like having MO.)


    @Andrew, funnily, my only proposal on meta.SE so far is about letting people comment and vote using reputation from a different SE site. Admittedly, that particular post isn't clearly written, but I still stand by that idea: there is a clear need for users of SE sites to be able to go directly there and participate in reporting/voting.


    There are some arXiv subject areas that aren't quite appropriate as tags. I'm not sure what do to about them. For example, there are "math.CA - Classical Analysis and ODEs", "math.KT - K-Theory and Homology", and "math.RA - Rings and Algebras".

    I'm thinking of

    • changing [analysis] to [ca-classical-analysis], leaving [odes] alone,
    • changing [k-theory] to [kt-k-theory], leaving [homology] alone,
    • changing [ring-theory] to [ra-rings-and-algebras] or [ra-ring-theory] and not having an [algebras] tag.



    How about [K-Theory] rather then [kt-k-theory] and similarly [Algebraic-Geometry], [Algebraic-Topology], [Rings-Algebras], [Classical-Analysis] etc.?


    @Ilya: I think you missed the part of the discussion where we were lamenting the fact that tags must be lowercase.


    @Anton, no, you misunderstand my proposal.

    First, it seems to be close to consensus that "ag-algebraic-geometry" tags are unpleasant. To summarize what people think about them:

    Ben "I'll admit, I'm not impressed."

    Andrew Stacey "@Ben: Okay, okay, I surrender ..."

    me "the topic deserves a top-level post and some thought going into it"

    But anyway, that's done. However, here's my suggestion: once we have the ability for uppercase letters in the tags, rename them to "Algebraic-Geometry" and so on, rather then "AG-algebraic-geometry" (which is redundant).


    @Ilya: you seem to have misunderstood Andrew's reply. We were having a joking exchange about my creation of the "field medal" tag.

    Apologies for the orthogonality, but can we please have a [finite-fields] and/or [characteristic-p] tags? Is there a place to request tags?
    @Sonia.B You can create new tags with sufficient reputation, and retag old questions with slightly less. (That said, I've just retagged a few with [finite-fields].) This thread is a good place to suggest mass retagging, so if there's something you think should be replaced by [finite-fields] or [characteristic-p], let us know!

    Actually, I'd prefer we keep specific tag requests in the Tags to Merge thread, and use this one to discuss general tagging policy (like what to do with arxiv tags)

    Perhaps it would help to rename this thread "Tagging Policy", and the other one "Tag Requests" or something similar?

    Done. By the way, if you want somebody to create a new tag, like [my-new-tag], you should ask directly on MO. For example, if you find (or post) a question that should really have the [my-new-tag] tag, leave a comment to the effect of "could somebody please retag this with [my-new-tag]" and a 500+ rep user will probably take care of it within an hour.

    @Scott: It's the other way round: Retagging requires more reputation (500) than creating new tags (250). Unless the FAQ is wrong, of course.

    Greg Kuperberg has suggested the naming convention [ag.algebraic-geometry] for arXiv tags. Given that we can't do capital letters, I kind of like it.

    I'm in favour of the [ag.algebraic-geometry] tagging scheme, although I still prefer [], especially if capitalisation is one day allowed.