12

Prologue: I am going to be using the Theoretical Computer Science as an example. I posted my question here because I believe my question is of a general nature. I am not picking on TCS in any way.


Tag taxonomy consumes a great deal of time on Meta.StackOverflow, particularly when dealing with Stack Overflow specifically. In the process of asking my question, I hope the reasons why become clear.

One of the things I have the most difficulty with when I visit a new Stack Exchange site is figuring out how the tag system works on that particular site. In my experience, I have found that every site has radically different philosophies with respect to tagging, so your tagging skills are not transferable from site to site. I still have this problem on Meta; in fact, you will see that someone will come along shortly and fix my tags for me, on this very question.

Today I stumbled into a question on Theoretical Computer Science, and noticed that the question I was looking at was tagged cc.complexity-theory. I wondered, "How does that work? What does cc mean? So I went to the Help Center, and found the Tagging article, but it doesn't say anything about these prefixes.

I then put [tags] in the search box on their Meta site, but didn't find anything that stood out. I thought, "Well, these are post-graduate types; I bet they have their own FAQ." Sure enough, it is here. That post led me to this answer, which immediately hits me with the term, "arXiv tags."

Hmm. What's an arXiv tag?

Right below that, there's an inscription that says: "[FIXME: add answer]". OK, I guess it's under construction. Well, it turns out that there's a supplemental FAQ with a similar "How do I tag" link in it, except this one links to

arXiv tags vs not?

Well, OK, that link was also right below the "Under Construction" sign. In any case, this post describes what arXiv tags are for, but doesn't establish any community consensus. If anything, there's mostly dissent there.

And I still don't know what an arXiv tag is.

Oh Harold... Do you see what I'm getting at?

So here's my question. How are new users expected to know how to properly tag questions? What is the canonical resource that we can steer them to that will help them answer the question: "How do we properly tag on this site?

20
  • hi RH. not picking on tcs.se? you should! arXiv is well known to scientists. it has its own subj classification. wasnt there a kaggle contest for an AI algorithm to suggest tags on posts? that will be a great feature if ever implemented. re tags it seems that there is a lot of confusion on stackexchange incl by mods about the nature of a folksonomy ... mods tend to "police" the tags and try to kill tags they dislike, really dislike this attitude on the collective intelligence aspects of tagging. note however the tag search works well on the sites. – vzn Feb 8 '14 at 2:46
  • here it is kaggle se tag prediction contest – vzn Feb 8 '14 at 2:50
  • MathOverflow also uses arXiv tags, which causes problems when migrating to Math SE. – TRiG Feb 8 '14 at 2:50
  • @vzn: Meh, was expecting a link to modabuse, not Wikipedia. I am disappoint. – user102937 Feb 8 '14 at 3:28
  • 2
    @TRiGisTimothyRichardGreen: Actually, that's no longer the case. See this meta.MO post, and the associated meta.math post. – user642796 Feb 8 '14 at 4:59
  • To avoid confusion for those not clicking through the links given by @ArthurFischer: it is no longer the case that it causes problems for migration (but it is still the case that MO uses them) as synonyms got set up on target sites. Actually, I would say that these tags existed in the end is helpful related to migration. – quid Feb 8 '14 at 11:25
  • Having had a look at the relevant meta it is a bit unclear to me what problem there is. Okay the FAQs are not yet complete, but the linked to discussion make clear (though perhaps somewhat indirectly) what an "arXiv tag" is. For instance see meta.cstheory.stackexchange.com/a/55 , note especially "Here is the list of ArXiv/CoRR subject classes that I have used as a reference." – quid Feb 8 '14 at 11:51
  • @quid: [sidebar, since I still think this post is applicable generally] Why has a community consensus not emerged on the use of arXiv tags? The top voted answer on that post seems to suggest that they not be used at all. Have they been forced onto the community? – user102937 Feb 8 '14 at 18:43
  • I think there are two different questions you seem to be conflating: a) should a user use an arXiv tag when asking a question. b) should the tagging system of the site be organised around some top-level tags, and b') if so should they be derived from relevant arXiv categories (a large preprint server for articles in various sciences) and b'') how should, if at all, these top-level tags be distinguished from the rest. Whatever the discussions in 2010, the site adopted this tagging-system (as it is in their FAQs and the tags get a lot of usage) thus as a (casual) user you should use them. – quid Feb 8 '14 at 19:51
  • If you think the description what is an arXiv tag is not clear enough there, then I would say that you have a point, and the way to change this is to ask on the sites meta (or to comment on the relevant FAQ entry or something). If you think the tagging system in general is flawed then also bring it up on the sites meta (but rather not if you are not a regular user). On the general point I think the default place is/should be the sites meta-FAQ, which in this cases actually contains the information you want. Sure, it could be improved, but if you ask for this there I guess it would happen. – quid Feb 8 '14 at 19:55
  • @quid: I think my point is that the post that describes your tagging system is three steps removed from the help center, and when you get there, the top answer says "I don't think we should be using these tags at all." That is the specific problem with TCS. The more general problem is that this kind of tag usage information is not readily available in a "well-known" place, anywhere on the network. – user102937 Feb 8 '14 at 20:17
  • Yes I agree it is not optimal in general; I might write an answer later. Just for completenes, the "your" is out of place, I do not even have an account on TCS. On MO things are also still improvable, but the change into the network caused a lot of things to be done; but see meta.mathoverflow.net/questions/1467/… for how it used to be. – quid Feb 8 '14 at 20:28
  • 1
    While you make a fair point, the disclaimer in the beginning of your question cannot be taken seriously: you very much are picking on CSTheory, and your complaint is very CSTheory specific. Why don't you just post on CSTheory meta, and hopefully the faq can be clarified. It is not an issue we've had problems with, because most of the people capable of asking a research-level question (everything else is off-topic) would know very well what arxiv is, and what the arxiv classification tags are. – Sasho Nikolov Feb 8 '14 at 22:12
  • @SashoNikolov: I don't particularly care about TSC. The folks there are probably smart enough to figure it out anyway. I used TSC as an example because it brings a network-wide problem into sharp relief. – user102937 Feb 8 '14 at 22:34
  • Does this extend to meta sites? Seems to be a lot of confusion over when to tag e.g. "bug" vs "feature request". – Troyen Feb 8 '14 at 23:02
6

I have long struggled with tagging. I think I am OK with it now, but it is generally something that I struggle with on many sites when asking a question.

This is also something I see frequently on meta. Frequently the only tag that you see on a meta post from a new user is one of the required tags and nothing else.

This is not a perfect solution since it is not tied to the help center, but might be a good first step for most site. I think one of the first things a site needs to do after graduating from private beta (maybe around the same time the pro tem mods are named) is to establish a "How to Tag" meta post.

The idea at the start would be to allow the community to flesh out and come to some sort of consensus on their tagging methodology, and then it can be turned into a once they get close to a consensus. The fact this is a meta post would allow the idea to be easily updated as the community matures and the philosophy on tagging changes. This could start with a single post, but could expand to multiple as the tags expand and the site grows.

The post(s) should address:

  • How to select tags: this is the canonical resource you would be looking for new users to understand the tag taxonomy of the site.
  • How to find the tags you are looking for: sometimes this is very easy, but sometimes it isn't, usually because of how a tag was named (see next bullet), but some clarification on how to identify what tags you might want to include.
  • How to name tags: naming consistency is one of the biggest tag issues I've noticed in some sites, especially when to abbreviate, whether or not to use a hyphen, when to use plurals. By establishing what you prefer initially, you can make it easier over the long term to maintain tags and fix the tags that don't comply with the community preferences.
3

I think you have a point and we should fix the issue so users who are not familiar with arXiv tagging can figure out it easily.

Let me explain why it hasn't been fixed yet. Part of the problem it is not fixed so far is that our faq has not been updated since the roll out of the new help center, part of it because the moderators cannot edit help center pages except one, and part of it is that this has not been fixed is that no one have complained about it to us over the past 3 years :) i.e. for the target community of the site these tags work without explanation: arXiv is the main online repository for computer science preprints and any researcher in computer science would be familiar with it. And even if someone is not familiar with arXiv tags when they type their intended tag they typically end up with the corresponding arXiv tag, e.g. if you type complexity-theory you will automatically get cc.complexity-theory with tag wiki right below it (we have tried to make the tag wiki for these main tags well written).

By the way, there was a consensus about using arxive tags on cstheory, we discussed it extensively and decided to use it since it is most useful for the target community of the site. Any researcher in CS would be familiar with them. This is partly modeled after MathOverflow's tagging system. We think it is good to have two kinds of tags: one for categories and one for keywords and we would like these category tags to be distinguishable. There is a well structured way of dividing topics in sciences into categories (see also ACM Computing Classification System or AMS Mathematics Subject Classification) and these help our users find and filter the questions they are interested in more easily so we built our tagging systems on these.

I think there is often a trade off between the ease of use of SE sites for users in the target community of a site and users from other SE sites. We think of our sites (even if they are relatively small compared to the trilogy) as sites for themselves not side projects of trilogy, so we focus more on the needs of the communities of each site as they are the people who make these sites work. As the number of SE sites grows it is expected that there will be more diversity in how different SE sites work. I think hard science sites are still rather close to the trilogy because of their nature, but sites for a soft topic in humanistic or arts (e.g. philosophy) the policies they would require will be quite different from those that the trilogy sites require. But of course we should also make them as accessible as possible for users outside each site's community. E.g. one thing we should have done in this case is to set a synonym from complexity-theory to cc.complexity-theory so if someone is not familiar with arXiv tags and used the first one for search it would automatically get converted to the correct tag on the site.

2
  • I could be wrong, I don't think this topic is specifically about arXiv tags. It seems like the general point of this discussion is to say "hey, SE needs to add a community-specific page somewhere where sites can explain their tagging system." – Troyen Feb 8 '14 at 23:10
  • @Troyen, I understand that Robert's point is more general. I agree that we need a community mod editable help page for explaining the tagging system for each site. However it was Robert's main example so I tried to explain how we end up in here with the hope that it might help. – Kaveh Feb 8 '14 at 23:17
3

A natural place for some quick information on the sites tagging system seems to be the /ask page. We also have many users that are new to all this, often they will not look through the Help Center, and almost certainly they will not know they could look in meta for some FAQs. But they must see the /ask page before asking.

In fact, we had this on MathOverflow (before we joined the network) it looked like this:

Old Tag Box

(also see the pending feature-request on meta.MO Tag suggestions on ask page to get this back).

This did not convey everything about tagging on MO, but still the most relevant information was there, namely that there are some tags that are essentially considered as required tags, what they are, and some additional information about them (via the link that went to http://arxiv.org/archive/math if memory serves well, but in any case something comparable)

A solution to the problem raised in OP could be that a site-moderator editable box is inserted on /ask where the site's moderators could give some information on tagging for the specific site (which could include links to relevant meta threads if they so choose). This same information could also be included in the Help Center page on tagging (which at the moment AFAIK by default does not have the option of 'per site' customization).

You must log in to answer this question.