While cleaning up tags, I noticed that we could make better use of some computer science arXiv categories to supplement the math arXiv tags that we already use.

The most relevant categories appear to be the following:

• cs.CC - Computational Complexity - Covers models of computation, complexity classes, structural complexity, complexity tradeoffs, upper and lower bounds. Roughly includes material in ACM Subject Classes F.1 (computation by abstract devices), F.2.3 (tradeoffs among complexity measures), and F.4.3 (formal languages), although some material in formal languages may be more appropriate for Logic in Computer Science. Some material in F.2.1 and F.2.2, may also be appropriate here, but is more likely to have Data Structures and Algorithms as the primary subject area.
• cs.CG - Computational Geometry - Roughly includes material in ACM Subject Classes I.3.5 and F.2.2.
• cs.GT - Computer Science and Game Theory - Covers all theoretical and applied aspects at the intersection of computer science and game theory, including work in mechanism design, learning in games (which may overlap with Learning), foundations of agent modeling in games (which may overlap with Multiagent systems), coordination, specification and formal methods for non-cooperative computational environments. The area also deals with applications of game theory to areas such as electronic commerce.
• cs.CR - Cryptography and Security - Covers all areas of cryptography and security including authentication, public key cryptosytems, proof-carrying code, etc. Roughly includes material in ACM Subject Classes D.4.6 and E.3.
• cs.DS - Data Structures and Algorithms - Covers data structures and analysis of algorithms. Roughly includes material in ACM Subject Classes E.1, E.2, F.2.1, and F.2.2.
• cs.DM - Discrete Mathematics - Covers combinatorics, graph theory, applications of probability. Roughly includes material in ACM Subject Classes G.2 and G.3.
• cs.FL - Formal Languages and Automata Theory - Covers automata theory, formal language theory, grammars, and combinatorics on words. This roughly corresponds to ACM Subject Classes F.1.1, and F.4.3. Papers dealing with computational complexity should go to cs.CC; papers dealing with logic should go to cs.LO.
• cs.IT - Information Theory - Covers theoretical and experimental aspects of information theory and coding. Includes material in ACM Subject Class E.4 and intersects with H.1.1.
• cs.LG - Learning - Covers machine learning and computational (PAC) learning. Roughly includes material in ACM Subject Class I.2.6.
• cs.LO - Logic in Computer Science - Covers all aspects of logic in computer science, including finite model theory, logics of programs, modal logic, and program verification. Programming language semantics should have Programming Languages as the primary subject area. Roughly includes material in ACM Subject Classes D.2.4, F.3.1, F.4.0, F.4.1, and F.4.2; some material in F.4.3 (formal languages) may also be appropriate here, although Computational Complexity is typically the more appropriate subject area.
• cs.MS - Mathematical Software - Roughly includes material in ACM Subject Class G.4.
• cs.NA - Numerical Analysis - Roughly includes material in ACM Subject Class G.1.
• cs.SC - Symbolic Computation - Roughly includes material in ACM Subject Class I.1.

The categories cs.AI, cs.CL, cs.CE, cs.PL could be relevant to mathematics, but we don't seem to have a significant number of MO questions in these categories.

[...]

The new tags [cs.cr.cryptography], [cs.ms.math-software], [cs.sc.symbolic-computing] have been created. We're waiting a bit for comments on cs.FL since merging two tags is irreversible.

We don't have [math.AT.algebraic-topology], for instance...

Dear Joel,

The old tags [cryptography], [mathematical-software], [symbolic-computation] no longer exist, they have been renamed [cs.cr.cryptography], [cs.ms.math-software], [cs.sc.symbolic-computing]. This is just a trial and this is completely reversible. (The [formal-languages] and [automata-theory] tags have not been touched since that would be irreversible.)

The reasons you point out are exactly why I posted this. As I mentioned above, very few game theory questions have a cs angle, so it would be inappropriate to rename [game-theory] to [cs.gt.game-theory]. A similar argument applies to logic, in fact the descriptions of math.LO and cs.LO are so different that it wouldn't be inappropriate at all for [lo.logic] and [cs.lo.logic] to coexist. For cryptography, mathematical software, symbolic computation, the descriptions of the cs categories seem to agree with how the community uses these tags. For formal languages and automata, the main issue is that these are conceptually very different though one hardly ever goes without the other.

I find your objection to the cs prefix interesting. I'm glad you mentioned it. In my mind, all MO tags have an implicit math prefix because all MO questions are primarily about mathematics. (Questions that are primarily about cs should be redirected to cstheory.stackexchange.com or cs.stackexchange.com.) To me, the cs prefix only indicates that the cs angle is relevant. If most users think that the cs prefix indicates that the topic is primarily cs, then I see that as a good reason to eliminate all cs prefixes!

That's a good point, Joel. We'll wait a few days for people to voice their opinions...

If we decide to eliminate cs arxiv prefixes, then we will go forward with the following tag renamings:

• [cs.cc.complexity-theory] -> [computational-complexity]
• [cs.cg.comp-geometry] -> [computational-geometry]
• [cs.lg.learning] -> [learning-theory]
• [cs.cr.cyptography] -> [cryptography]
• [cs.ms.math-software] -> [mathematical-software]
• [cs.sc.symbolic-computing] -> [symbolic-computation]

I second what Kaveh says in the second paragraph.

I don't understand what role the two-letter prefixes play in any of the tags. For example, why not change "nt.number-theory" to "number-theory"? The "NT" in the arXiv is just a space-saving abbreviation for "number theory" anyway, and it seems silly to use the abbreviation at all if you're always going to append the full name. Right now, the only role the prefixes seem to play is as a mildly obscure way of emphasizing that these are the same categories as on the arXiv. However, with the exception of ho.history-overview (where the usage of the category on the arXiv is broader than one might expect from the name), I don't think see any real content here. It's not like someone's going to say "Oh, I get it, you mean number theory as in the arXiv!"

Rereading the 2009 thread linked above, the basic idea was to have some reference list of "major tags" for classification purposes. (The term "major" here is used in opposition to "specialized" but not as broad as [analysis], [algebra], [geometry], or even [math] which do very little for classification purposes.) An ideal classification system would be such that every MO question would fit into one or more of these major categories, excluding off-topic questions. The consensus was that the math arxiv tags would serve that purpose well and they did for quite some time. It does appear that MO has outgrown this classification system.

This brings us to some basic questions:

1. Does MO still need major categories?
2. If so, what would be the list of major tags and how would it be maintained?
3. And how would one distinguish major tags from more specialized tags?

Note that, as a result of the 2009 discussion, a list of arXiv tags was added at the end of the ask form, and the front page has this notice: "Want to help? Consider retagging questions with no arXiv tag."

(I edited the title since this the topic is shifting to arXiv tags in general.)

One of the problem that is nicely addressed with the arXiv tags is that it solves the problem of the tags being not otherwise defined (which can be problematic coupled with the character-number limit in the tag names; compare [ca.analysis-odes] to [pseudodifferential-oper]--incidentally this also gives a partial counterexample to Henry Cohn's assertion that we almost append the full name).

Another particular strength of the arXiv tags is that the list is remarkably short. Imagine requiring using something like the MSC2010 for the tags! So for question 1, I agree with Kaveh that there needs to be a small number of major tags.

+1 Gjergji.

Francois, I should say that tag renaming and deletion abilities are considerably reduced in SE 2.0, at least for "local" moderators. If there is going to be a migration at some point, I suppose tag cleanup should be done before that.

The first thing I know is that it is not longer possible to delete tags (without retagging).

I also believe that renaming also suffers from this issue (we had a recent decision to change [filter] to [filters] and we had to do it manually, despite moderators being involved, it ).

I'm sure that one of the MSE moderators that writes here could be more helpful, but I do recall that some privileges were taken away.

@François, Can you make [cc.complexity-theory] a synonym for [computational-complexity]? Similarly [cg.comp-geometry] -> [computational-geometry] and [cr.crypto] -> [cryptography].