When should character tags be created/recreated?

Question

We've recently been going through and cleaning up tags - particularly character tags.

One of the criteria we've used to determine whether or not we should get rid of it, is to consider how many questions have been tagged with it.

Judging based on this measure seems like a sort of self-fulfilling curse though - getting rid of a tag entirely because there's not enough usage of the tag, and removing the tag from incoming questions prevents the tag from ever having enough questions tagged with it to warrant the existence of the tag.

Let's say we kill the tag due to lack of usage, and then continue to remove it from incoming questions - at what point do we say, "OK, we now have enough questions about this character to warrant a character tag?" Do we use quantities of questions at all?

If we re-create the tag at this point, do we go back and re-tag the relevant questions? If not, a zealous user could see it as a tag not being used enough, and wipe it out again.

Certainly, not every character needs or deserves a tag, but there are a number who do.

What guidelines should we follow regarding the creation, deletion, or re-creation of character tags?

Answer 1

Warning: this is a long one, mostly review! Skip to the end for the good stuff if you're in a hurry.

---

I've gone back and read a lot of the earlier questions, and the main problem is they were written when the site was much younger (the main "reference guide" is from 2011), and there were far fewer questions. I think it's worth revisiting how to apply those ideas to our current situation.

First, some source questions, newest to oldest:

• Please, remove "yoda", "darth-vader" and "obi-wan-kenobi" tags
• Why is there a voldemort tag? Is there a criterion for character tags?
• Inconsistent tagging
• Star-Wars character tags

If you go by those meta discussions, character tags have always been frowned on; obviously, that's not a policy that was strictly enforced, or we wouldn't have so many, but at least it has been consistently stated.

First off, the point of tags in general, as quoted from @Gilles' answer to the original Star Wars question:

Tags are useful for two things:

For people to subscribe to them or ignore them.

To classify questions in a manner that goes beyond plain text search.

He goes on to explain why character tags typically don't fit either of those two criteria:

For example a Star Wars fan would want to browse every question tagged star-wars, and a Star Wars hater would want to ignore these questions. But are you going to ignore questions tagged jar-jar just because you hate the character? If you're in love with Princess Leia, does that mean you specifically want to read all princess-leia questions and not other star-wars questions?

A question about R2-D2 would contain “R2-D2” or at least “R2D2”, so the tag isn't helping much.

In a comment to that same answer, @DavRob60 first mentions one criterion for determining if a character tag is a good tag:

when we will get more star-wars question, and a rather large subset of them on some specific character, the need of character specific Tags would arise.

@Jeff also seems to echo this sentiment in his own answer

I'm uncertain that there will be a great need for them - do these characters really have enough unique questions to warrant their own tags?

Both of these ideas pop up again in subsequent character tag questions, such as this answer (again from @Gilles) about voldemort:

A tag for a character is only warranted if there are a lot of questions about that particular universe (HP and ST do qualify), and a significant fraction of these questions are specifically about that character. If the tag is used on every question that mentions the character, that defeats the purpose.

Finally, as this topic has come up again recently, @phantom42 brings up another thing to factor in (again, initially related to Star Wars):

Given that there are entire solo books and comics dedicated to vader and obi-wan, I'm inclined to let those ones live. I think the amount of dedicated material should play into our culling

---

Current Guidelines

Which brings us to this question, and what our current guidelines should be. In trying to boil down all the past discussions to use as guidance for our cleanup efforts, I think we've settled on the following general rule for character tags:

A character-specific tag is only useful, and should only be created/kept, if there is enough material unique to that character, within their larger universe, to generate questions specifically about that character, such that one or more of the following is true:

• Users may want to follow/ignore just that character, but not the rest of the work.
• Someone may be an "expert" on just that character, and no other part of the work.
• The character appears in a significant amount of their own material

So far, some of the metrics we have been informally using, based on coordinating our efforts in chat, have been:

• A character that has their own spin-off titles automatically gets to stay.
• A character tag with few questions and no followers is an easy delete.
• A character that has lot of questions where they're the only character tagged in it, that tag will likely stay.
  • e.g. if chewbacca only ever appears with han-solo, we say goodbye to Chewie.
• A character tag with lots of people following it probably stays.

But that's obviously an incomplete list, so it can definitely use enhancement and refinement.

(on a side note: all the old meta discussions seem to agree that character tags are bad unless there's compelling proof otherwise. Assuming that's still the consensus, the criteria we come up with ought to be pretty hard to meet.)

Answer 2

First of all, i'd like to stress that I agree with a lot of Mike said on his answer. This answer is largely a rebuttal of ONE very specific point he references in his answer - but, to me, that point is crucial in determining whether the concept of character tags is worth defending or not; whether character tags as a group are useful or not (leaving aside choosing which specific character is or isn't worth a tag).

Mike quoted Gilles' point thusly:

Tags are useful for two things:

2. To classify questions in a manner that goes beyond plain text search.

And goes on to quote Gilleses theoretical reasoning for rejecting #2.

A question about R2-D2 would contain “R2-D2” or at least “R2D2”, so the tag isn't helping much.

The problem is, I don't know what Gilles's reason for stating this was (e.g., his experince with specific searches at the time this was written; OR simply a theoretical assumption), but in practice, I personally found Gilles' assertion in the second quote to be 100%, completely, incorrect (_{with all my respect for Gilles' encyclopedic knowledge, he does sometimes make mistakes too}).

• Personal testimony:

  The character tags significantly improve the efficiency and reduce the effort of searches for me. In my experience, you frequently have 3-4 times more posts that merelty mention a character name, compared to posts specifically about that character. I've literally had cases where I spent 30min searching for something with no success, only to think of using a correct tag in the search, and a minute later have 10-result thing with what I was looking for among those 10.

• Mistake in Gilles' own reasoning

  He himself admits that there's TWO searches one needs to do, "R2-D2" and separatelty, "R2D2". Oh, and he forgot "artoo-deetoo" or whatever other spellout of R2's name both the EU writers and site users choose.

  Since the site search doesn't allow complex logical queries; that means the user now has to repeat their search at least 4 times. And might STILL miss what they need because they forgot about searching for "luke's astromech droid" because someone didn't bother putting the droid name in the question text, or "artoo" because someone didn't bother with full spellout.

• Evidence

  Let's look at an example. Since R2-D2 tag was actually vetoed and is I think on like 1 question, let's look at another tag that aroused Meta conversation: He-who-must-not-be-tagged

  • This is the search using voldemort (_{I'm adding "HP" tag for a technical reason related to search mechanics on SE, it's not really needed as far the result counts}):

    "[harry-potter] [voldemort]": 250 questions. A lot, but manageable to traverse.

  • Now, let's pretend I need to do my research but the SFF.SE community reached a consensus, and Harry-Pottered that tag:

    "[harry-potter] voldemort": 3,042 results . Personally, at this point I give up and admit I'm not star-trek-data, I just can't scan this.

  • OK, the above was the worst-case scenario, assuming I'm a "normal" user and wasn't aware that you can only search question text. I hope nobody will argue that most users are aware of "is:question" syntax?

    But let's say a very experienced rare user who knows this syntax tries:

    "[harry-potter] voldemort is:question" : 784 results

    Yes, it's not as astronomically bad as 3000, but still far more than a normal person would even remotely would be willing to peruse for research.

    And as I stated above, this makes an extremely unlikely assumption that our searcher is even aware of this syntax.

    • To make things worse, our text search missed 17 questions mentioning "Voldermort", 4 of which would NOT have shown up in the original searches

The reason for this major difference is that Voldemort may frequently be mentioned in questions or answers on Harry Potter, but precious few of those mentions (< 1/3) are ABOUT Voldemort.