Tagging: are we doing it wrong?

Question

I wanted to post this for a while...

First off, I think that we're not tagging enough. Currently, we have an average of 2.1 tags per question. I've checked some (thematically) comparable sites, and Stack Overflow, Code Review and Programmers all have between 2.6 and 3 tags on average. Tags are a really useful concept of SE to organise the questions on the site and make it easier for people to find those they are interested in, and I think we should make use of that. I tend to do quite a lot of tag edits on new questions if I know that more applicable tags exist, but it would be nice if I could get some help with that.

In addition, while we've added a few tags for challenge types recently, it seems that some people put more effort into getting rid of tags than adding new ones. As far as I'm aware I'm one of very few people who regularly try to add new (non-challenge) tags.

At the same time, we do have a few tags that are completely useless, but still widely used. Luckily, we got rid of [algorithm] recently, but there's still stuff like printable-ascii, which I think is the epitome of a meta tag.

And then there's really confusing groups of tags, like math number arithmetic, where I'm never quite sure which ones to apply to any given maths-based challenge.

So before we embark on The Great Tagging Spree, I think it would be useful if we could get a consensus on what kinds of tags we actually want. The normal definition of a meta tag ("The tag can't stand on its own") does not really apply here, since challenge tags are mandatory. Can we come up with some guidelines for how non-challenge tags should be applied and which ones we should even create or keep?

Answer 1

The question I wanted to set out to answer was — what are we doing wrong? Why do we have bad tags, and end up needing to purge them?

To answer this, I went through the top 100 tags or so and tried to categorise them:

Winning criteria

code-golfcode-challengepopularity-contest fastest-codeking-of-the-hillatomic-code-golfrosetta-stonefastest-algorithmcode-bowling

Input/output types

ascii-artstringnumberimage-processinggraphical-outputarray-manipulationdateaudioanimation

Restrictions

restricted-sourcequinesource-layoutprintable-asciipolyglot

Language-specific tags^*

cpythonjavascriptc++brainfuckc#assemblyphpcssmanufactoriaregular-expressionlogic-gates

Task types

tipskolmogorov-complexitycompressionai-playerinterpreterparsingmetagolfpuzzle-solvercode-generationoptimizationgenerationconversionprogramming-puzzlebusy-beaveroptimized-outputcountingsimulation

Themes

matharithmeticprimescombinatoricsnumber-theorygeometrysequencesubsequencefractalpermutationsbase-conversioncomplex-numbersprobability-theoryset-theorygraphssortingcryptographynatural-languagefile-systemdata-structurescellular-automatahashingfunctiongameboard-gamerandommusicwordchemistryword-puzzlechesssudokucrosswordtic-tac-toemazetetrishello-worldpalindromepibinaryfibonaccidivisiongame-of-liferoman-numeralsmorsenew-yearskeyboardbitwisebinary-tree

^{*This includes questions which require answers written in the language, as well as questions using the language as a theme}

(I had trouble distinguishing between the last two categories...)

---

That gives us an idea as to what classes of tags we might want, but doesn't directly give us a way of immediately telling if a tag is bad. But from browsing the tags above, I think the biggest problem is that there's a lot of confused tags, which aren't clear about what they're meant to cover:

• printable-ascii simultaneously describes challenges where the code may only contain printable ASCII, the input is printable ASCII and... the output is printable ASCII?.

• function has questions to do with calling functions, questions where code must be a function and questions about mathematical functions.

• counting has questions about counting items/counting possibilities... and then stuff like this.

• animation is all okay except One Line Aquarium (again), which seems out of place. The tag wiki isn't clear about whether this tag is appropriate or inappropriate for this question.

• And don't even get me started on word-puzzle...

So my proposal is that tags of the above classes are fine, except:

• The tag name should be unambiguous enough so that people can get a good idea of what the tag encompasses, while preventing any possible major causes of confusion.
• The tag wiki should clear up any possible minor confusion, and should clearly define the scope of the tag (especially for winning criteria, restrictions and task types). To be honest, I think we would benefit from more detailed tag wikis, in general.
• The tag should ideally be broad enough to be applicable to at least a few questions, but not be too broad to the point where it's better to split it up into several smaller tags.

Answer 2

This answer is to suggest a policy for what we should consider an admissible tag. It doesn't really address the question of what to do with overlapping tags, like the maths family.

As I said in the question, I don't think the guideline "can the tag stand on its own" is very useful around here on in general. I believe I've seen Grace Note or Monica Cellio post on some early beta site, that the real question should rather be whether anyone would be interested in filtering by that tag - because that's what they are for after all.

So I suggest this guideline:

Is the tag useful for filtering questions? Would anyone really want to see all challenges with that particular tag, because they're interested in the kind of challenge implied by the tag?

As an example, I personally think that this would classify printable-ascii as a bad tag. It describes a fairly random property of the challenge, that doesn't really tell you at all what the challenge is actually about. And I don't see anyone thinking "oh I feel like answering a printable ASCII challenge today". If they were interested in restricted source challenges, they would filter by that tag - and more importantly only 3 of the 35 printable-ascii challenges are even meant to be restricted-source challenges. Most of the time, this is just a random feature of the challenge that isn't even intended to really affect the answers, but a mere technically (because of source layout etc.).

Please let your vote and comments indicate your opinion of the definition, not this example. Whether printable-ascii is ultimately a bad tag or not is a discussion for another day.