What is meta? ×
Meta Stack Exchange is where users like you discuss bugs, features, and support issues that affect the software powering all 150 Stack Exchange communities.

Single-use tags automatically expire after a few months. This is arguably the right thing when the tag is a misspelling (though I'd prefer some way of reviewing the process — but this post is not about that). However, if the tag was clearly deliberate, the default should be not to delete it. I propose the following approximation:

If someone bothered to write a wiki for the tag, do not delete it.

I posted this after browsing the orphan tag wikis on Science Fiction & Fantasy, most of which are about authors about whom questions have been slow to come. Some of these have been recreated in the meantime, with more than one occurrence even. I'm sure other sites have plenty of similar legitimate low-use tags. Saving the ones that have tag wikis would be a good step in the right direction.

You're DIFFERENT and That's SUPER

To clarify, this is about tags that have exactly one question: they get pruned automatically after 6 months, leaving questions with one less tag or untagged.

share|improve this question
I'd be up for this. It'd help on Gaming for sure, we have LOTS of single use tags for game titles that are NOT junk and should, ideally, not expire. – Ben Brocka Mar 31 '12 at 23:44
@RobertHarvey This is about tags that have exactly one question: they get pruned automatically after 6 months, leaving questions with one less tag or untagged. – Gilles Apr 1 '12 at 15:40
You can't get another question on the same tag within a six month period? YAGNI. – Robert Harvey Apr 1 '12 at 15:40
@RobertHarvey On the contrary, it's the single-use tag deleter I find YAGNI. The one question does show a need for the tag. It's not always about the most popular authors. – Gilles Apr 1 '12 at 17:24
@RobertHarvey: It's more of a different needs for different sites kind of thing. Remember: the point of deleting 1-question tags is handling a typo. On SO, a one-question tag in 6 months is almost always a misspelling, so much so that the few false positives aren't a problem. On most other sites, the number of false positives increases. Being able to have a tag wiki prevent the removal of a 1-question tag is a reasonable middle-ground. It allows a human being to say, "I've vetted this tag, and it is certainly a real tag that could be useful in the future." – Nicol Bolas Apr 1 '12 at 18:17
@nicol maybe, but it's also wishful thinking and speculation. I'm sympathetic to the idea that someone cared enough to write a tag wiki, but then if the tag can't muster two questions over a six month period, do you really need that tag? If you feel so strongly about it, don't bother writing the tag wiki, ask two questions on that topic yourself! That'll certainly be more visible and useful to the site than a tag wiki for 1 question, don't you think? – Jeff Atwood Jul 23 '12 at 0:08
@JeffAtwood Why is it useful to make questions about less popular topics (authors on Science Fiction & Fantasy, games on Arqade, programming languages on Stack Overflow, etc.) untagged after a few months? – Gilles Jul 23 '12 at 0:14
Wouldn't a better feature be to re-attach the tag wiki if the tag is ever re-created? I can only see this being abused. People will write crap tag wikis for crap tags just to make them stick around, and people will actually approve them just because approving is easier than rejecting. – animuson Jul 23 '12 at 0:19
@animuson That would be an improvement on the present situation, but it still lets known-good tags disappear for no good reason. – Gilles Jul 23 '12 at 0:24
@JeffAtwood: "consider it incentive to ask that second question about the topic." It's only an incentive if people know it's an incentive. And I don't think it incentivizes good behavior. Someone could just ask a second question that's marginal or something. Tags are for categorizing information; categories should not need to prove themselves to anyone. – Nicol Bolas Jul 23 '12 at 0:52
@Jeff: It sounds to me like you're trying to appropriate something for a purpose it was never intended for. The current system is to get rid of newly-created tags that are the result of misspelling; that's why it was invented. It was not devised as a means of forcing a tag to meet some minimum requirement before being allowed to exist. Nor was it devised as a means of effectively breaking the categorization of a question asked in good faith by making it untagged. – Nicol Bolas Jul 23 '12 at 1:59
This has happened to us on Biblical Hermeneutics, a small site still in beta. We have tags for individual biblical books. Last night I came across an untagged quesstion, tagged it, and wrote the tag wiki, all while thinking "didn't I write this once before?". Apparently yes; I didn't know until now that singleton tags get deleted (and their wikis go away with them). The deletion is a feature on SO but a bug on sites like BH. – Monica Cellio Oct 4 '12 at 21:50
@MonicaCellio Even on Stack Overflow, there are good tags that get lost now and then. It's difficult to keep track of them because the deletion is hard to audit since it rewrites history as if the tag had never edited. – Gilles Oct 4 '12 at 21:52
@JeffAtwood: And if we were talking about tags that do not matter, that might be something. But we're talking about tags that have tag wikis. It's a lot harder to say that they don't matter just because there's only one question for them. It obviously mattered enough to someone, as they took the time to write a wiki entry for it. This tool is catching more than users who "spray dozens of tags on every question". Indeed, I'd say it's poor at that, because it's very easy on SO for such tags to accumulate more than 1 question. – Nicol Bolas Nov 26 '12 at 2:01
@LanceRoberts There isn't much to say about this one. The implementation basically does what it says on the tin. This went live some days/weeks before I came across this meta post and realized we forgot to tag it. – Adam Lear Jul 21 '13 at 5:43

2 Answers 2

up vote 14 down vote accepted

Some sites have organized themselves around a primary tag categorization system. For a subject with a single, logical way to divide most questions, there will inevitably be hapax legomena. In fact, they will likely follow a Zipfian distribution unless the tag set is naturally finite and the question set is allowed to grow without limit.1 Unless your subject happens to fit that specific pattern and until you have enough questions to cover all the tags, you will find yourself duplicating effort retagging and rewriting tag wikis. So a feature designed to clean up the tagging system ends up breaking it!

Wikipedia mentions:

In the fields of computational linguistics and natural language processing (NLP), esp. corpus linguistics and machine-learned NLP, it is common to disregard hapax legomena (and sometimes other infrequent words), as they are likely to have little value for computational techniques. This disregard has the added benefit of significantly reducing the memory use of an application, since, by Zipf's law, many words are hapaxes.

It's efficient for the computer, but not for humans. Imagine if you were suddenly forced to communicate with the "1,000 most common and basic words in English". Imagine if you forced an expert to limit their vocabulary. Imagine if your local library pulled from the stacks all books that were the only representatives of their particular number in the Dewey Decimal System and tossed them in a pile in the lobby.

Since the tag silently disappears without a trace, it can be hard to restore the damage. If you happen to have ♦ privileges, you can recover the text of orphaned wikis, if you know where to look. But it's a hassle and a waste of time to fix tags that have been vandalised by the system. Meanwhile, simple misspellings are easily handled by merely creating a synonym. How is it that a clearly broken tag is handled more easily than an arguably good tag?

Of course there is a good reason not to allow singleton tags as a general rule: people are quirky. We've had tag suggestions for hades, nomadism, and bookend, which just aren't going to be things anyone will search for. But note that these are secondary tags in our system. Secondary tags are only useful if they cover a large number of questions. A person might come looking for textual-criticism questions, but who would ever look for nomadism?3

The point is, each site is (or ought to be) free to craft its own policies for question tagging. It's already possible to remove bad tags and even prevent them from being used again. But it's not possible to whitelist-community approved tags. The beauty of this suggestion is that it gives us an elegant way to define the limits of our site without being thwarted by the system's automated processes.


  1. The Bible has 66 books2 and there are no plans to expand the list. So if we keep asking questions, we will eventually get at least two per book. I'm not sure how long that will take, however, unless we make an effort to focus new questions on books that currently have just one.

  2. More or less.

  3. There was a question with that tag, which I thought was very good. However, the more general history tag works better.

share|improve this answer
I'm not a huge fan of speculative work, e.g. creating tags in advance of "needing" them, because "we're gonna need it". YAGNI. Questions and actions speak louder than speculation, so ask the questions on the relevant topics, or encourage others to. – Jeff Atwood Nov 26 '12 at 1:26
@Jeff Atwood: Please see Let's save our tags! Also note that deleting a singleton tag is a single button press for a sufficiently privileged user, but restoring an orphaned tag wiki requires detective work. I'm not advocating for tags (and wikis) to be created on spec, but I'd rather people waste their time doing that than undoing the damage the system does because it thinks it knows our subject matter better than we do. – Jon Ericson Nov 26 '12 at 5:43
well FWIW I think preventing single-question 6 month tag removal on the presence of a tag wiki is a reasonable middle ground. – Jeff Atwood Nov 26 '12 at 7:10

It's only an incentive if people know it's an incentive. And I don't think it incentivizes good behavior. Someone could just ask a second question that's marginal or something. Tags are for categorizing information; categories should not need to prove themselves to anyone

Categories indeed do have to prove themselves to everyone. Instead of, say, ten clearly defined tags, you could have ten thousand poorly defined, overlapping, inconsistently used tags. How is that helping the site, or the community on the site?

For example, that's why we eventually added the tag synonym systems, because we ended up with too many duplicate tags that meant the same thing. More duplicate tags meant more confusion -- which tag is the correct one? Which is used the most? Why do we have all these different tags?

That's also why we added the tag wikis so that people could understand the tags, what they mean, when to use them... ultimately, tags need to explain and justify their existence. And if they can't, they have no right to exist.

Before we had tag wikis, the only way for a tag to "explain" itself was in terms of which questions it was applied to. With only one question over a six month period, who can say what that tag means -- or if anyone even cares? It's just a bunch of litter cluttering up the tag system, and every user who types in a tag has to deal with potentially thousands of these orphan tags (we removed 7,400+ on Stack Overflow) matching what they type, and thinking.. "hmm, should I use this tag?"

Giving this a little more thought, it's probably an OK compromise to require the tag wiki to be completed if you want the tag to remain in place, even if there is only one question ever asked with that tag.

This will get rid of the accidental tags, the error tags, and the careless tags -- where people mindlessly spray a bunch of tags over a question because they can. The kinds of users who create poor single-use tags will not be motivated to create the tag wiki for that tag. And for those who feel strongly that a single question tag absolutely must exist, forever, they should be motivated enough to build a reasonable 300 character tag wiki for it if they feel so strongly.

Or, y'know, ask the second question on that topic in a six month period, if it's really so important.

share|improve this answer

You must log in to answer this question.

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