Automatic tag clean up and the long tail

Question

Stack Exchange sites have an automatic tag clean up system that removes tags with only one use after 6 months. However, it seems like this is detrimental to SciFi Stack Exchange, which has a long tail of questions on less popular / relatively unknown books or series (and surprisingly, some fairly well known ones too).

The result of this tag clean up system is questions being tagged with untagged because they no longer carry any tags. These can be easily cleaned up; what is more worrying is questions which once carried two tags now only carry one generic tag, making it difficult to discover questions from infrequently asked series.

In addition, there are tags which will probably see very few questions. For instance, this question is the only question on this site about the now cancelled TV series FlashForward, so it's unlikely any more flashforward question will be asked. These tags will have to be constantly recreated, since they are legitimate tags, and these questions do deserve the series tags.

So, my proposal is that this system be turned off for SciFi Stack Exchange.

---

PS. There seems to be an additional problem where the last editor of the question would appear to have introduced the untagged tag into questions, even though it is in fact the system that's doing the job. For example, in the revision history for this question, it might appear that Tony Meyer introduced the untagged tag, but looking at the Tweet you can see that it actually the now-deleted juliet-marillier that was added. This is probably a bug.

Answer 1

Developers can turn off this feature on a site-to-site basis. Arqade recently got rid of it. The request to turn it off on SF&F has been relayed to Stack Exchange staff.

---

badp has created a data explorer query that shows at-risk tags: tags that are used on a single question and have been around for at least 3 months. I reviewed the first 25:

• Required tags (name of the author or movie): john-steakley mercy-thomson eric-flint bridge-to-terabithia venture-bros jack-mckinney jean-m-auel godslayer land-of-the-lost the-giver the-prisoner emberverse clockwork-orange 1001-nights green-rider quiet-earth person-of-interest l-ron-hubbard k-pax mst3k fantastic-voyage

• Optional tags (name of the book where there is also an author name): armor-book painted-caves

• Tag that should have been different: quantum-thief (name of the title, we would normally use the author name — however this is a retag of a tag that had been deleted previously, so we can't know what was there befoere)

• Tag that makes sense, but isn't a great loss: aladdin (it's a concept tag, which we encourage, but it's a very searchable one, so it isn't particularly useful)

• Tags that we should get rid of: none

So that's 21 against the tag killer, 4 undecided, and 0 in favor. Kill the tag killer!

Answer 2

After reviewing Gilles' data and talking about this internally for a while, I agree that for a site like SciFi where there are no allowable classes of tags that apply to most questions, this culling isn't helping.

Disabled.

Please keep a sharp eye on https://scifi.stackexchange.com/tags?tab=new for crap tags.

Answer 3

It would be nice that instead of outright deleting a tag, it was either flagged for mod analysis, or better yet, for anyone with "moderator tools" usage, to decide what to do with it.

Answer 4

As an aside, I would create a separate meta post about the revision history making it look like the post author changed the tag to untagged. That looks like a bug to me.

For the main point: consider what the point of tags is. Despite what people often think, they aren't for search: it should always be simple to find a question you're looking for without the tags coming in to play at all (via the title, question, and answer content). Tags are all about following content (or the fairly useless semi-hiding content) that you're interested in.

For example, if I'm interested in doctor-who, then I follow that tag and those questions are highlighted (or, more usefully, I subscribe to the RSS feed for that tag).

If there's something that only has a single question within the tag-expiry period, then is it worth following? Perhaps: for example, I'm interested in following any questions about blakes-7, but for a long time there was only a single question, and so the tag was lost (until there was a second question, and the original question was edited to get the tag back).

However, I think this demonstrates that the current system works. While the site is small, those of us that read every question will generally notice when a tag can be reintroduced, and do so (this is where the untagged tag is useful, because you can easily see which questions are tag-less). When the site is large, it seems likely that everything you'd be interested in following would either have a tag already or would (in the case of new material) get multiple questions within the expiry period.

There's an easy solution if you really want a tag to stick around: ask another good question with that tag! Or two, or three! You can then follow it with confidence that it'll stick around.

I don't see how tags help discoverability at all. Surely people aren't skimming through the tail end of the tags page to find questions. There are lots of good ways to discover material from lesser known work that don't involve tags at all.

(As another aside: extending the expiry period, or maybe skipping it, for tags that have followers would be more worthwhile, IMO, since there's then evidence that they are of interest).