14

I wanted to get some help of others that may be able to leverage SEDE in ways that I can't.

I would like to see some analysis or statistics regarding how we actually, currently, are using tags.

When we're having discussions about tagging policies, I think it'd be beneficial to have some clear data on the subject that we can all work from. I'm not just asking about character tags, but all tags in general. I've an interest in this topic now because the term folksonomy was recently used, which I'd not heard of before. I'm trying to see how our folksonomy has developed.

Please post what you're able to find.


I was trying to build a SEDE query that someone else may be able to help me with or do. I wanted to pull a list of say, our top 50 tags, and then a list of the top 3 tags that are used with those top 50 tags. For example, one of the top tags is , what are the top 3 other tags used in those questions? Rinse and repeat.


Also, please see Tag Wars Episode I: Harry Potter and the Tag Cleanup

Which is an attempt to give certain tags a score, based on how they meet criteria, to demonstrate their currently quality/usefulness or lack thereof. A summary of those results, once they come in, would make an excellent addition to this thread.

  • 2
    I'll take a stab at this tonight. – user1027 Jan 4 '16 at 20:55
  • 5
    How do we use tags? Inconsistently and with great incompetence. – Valorum Jan 5 '16 at 11:26
  • Does it let you vote on this synonym? It wouldn't let me propose it without adding that tag to a question, and now I removed interesting-tags from my question again. (That's one of the default tags that the system won't clean up.) – Molag Bal May 18 '16 at 18:25
  • @anaranjada What's the point of this? – user31178 May 18 '16 at 18:26
  • I just thought interesting-tags should be a synonym of tags, since the former tag won't go away under any circumstances. And now I'm curious if the system will allow votes on the synonym. You can vote either way, obviously. Rand didn't like the proposed synonym, so you and he could kill it easily enough. (Also, I wanted to use my newfound 2.5k power...) – Molag Bal May 18 '16 at 18:27
20

The portion concerning naming conventions has been moved to a new answer, to make it easier to follow (and edit).


I've started off looking at the top tags, which I arbitrarily set as tags with count >= 50.

I just wanted to see what was there. I think the popular tags are the ones which the majority of users are going to be exposed to, and thus shape their ideas about how to tag and what kinds of tags to make.

In order to do this analysis, I had to come up with some way to categorize the tags. I wanted to keep to a small number that would include, theoretically, all of the tags.

Categories

Here's what I came up with:

  • Franchise - A name that generally refers to a franchise, which I define as a set of related works that span multiple series or multiple forms of media. The best examples are some of our top questions: , , , .
  • Series - An individually named series which may or may not be part of a franchise. See and
  • Work - An individually named single work, such as a movie or book, which may or may not be part of a series or franchise. See or
  • Work-specific - A character, setting, object, etc. that relates to a specific work, series, or franchise. See , ,
  • Author - Specific authors of written works
  • Publisher - The publisher of works. Only 2 examples are in the 50+ range, those being and , but smaller ones such as and are found.
  • Media - The format of the work in questios. See ,
  • Genre - Genre classifications of works, such as , , , and . Genre and Media classifications have some interesting uses, which I will point out in detail later.
  • Topic - Broad topics you can't necessarily pin down to a specific work or genre. See , ,
  • Probing - Tags used for questions used to clarify a work, its content, or its creation. The main example being our top tag, , but a large number of others exist such as , , .

Category analysis

After looking at these top 173 tags, I found that they broke down as follows:

By number of tags:
enter image description here

By number of questions tagged:
enter image description here

What shouldn't be surprising here is that Franchise tags are not only the most common types of tags, but the most commonly tagged questions. Part of this is because of how I classified franchise. What I did find surprising is that Work-specific tags are our 3rd most common type of tag, but 6th most commonly tagged questions. Although, Work-specific tags have more tagged questions than any 2 of the least commonly types of tagged questions: Publisher, Author, Work, Genre.

A complete breakdown of what I classified these tags as can be found in my other answer.


Relationships with Probing tags

I found Media and Genre tags to have some interesting things going on with them. They're rarely used alone, and tend to be attaching to some type of Probing tag. The most common example being , which is what I'll use in my examples.

Media tags:

Genre tags:

I included all of the decade tags, as only 2 more existed that weren't in the 50+ range. However, decade tags have since been burninated!

Combining both into one table, here's a quick chart showing the %age of each tag that is used with

enter image description here

The average percent of Media and Genre tags used with is 66.7%. That is, the average of the percents seen in the list and graph above.

The percent of Media and Genre tagged questions that include is 60.5%.

Media tags

Although seems to be an exception to the Probing connection, a cursory look at those questions shows that the tag is often used to specify that a question is only about the comic-book version of events or characters, instead of the entire franchise which may include movies and TV shows. This is how Media tags are used in general: they are almost never used alone, and make the most sense when used in conjuction with other tags to narrow scope.

Genre confusion

The two big genre tags, and have some counter-intuitive definitions based on actual usage:

Should not be used to categorize questions about specific works of science fiction. Should not be used to categorize questions about specific works of Fantasy.

However, none of the other genre tags have such a caveat. Furthermore, while this "rule" exists in the tag descriptions, it's not universally enforced. 53 questions include one of these two genre tags. I think it would actually be much higher, if not for tag-definition and enforcement, where those two genre tags are actively removed from Probing questions. I don't see any reason why a site named Science Fiction and Fantasy would have more questions tagged (x165) than either (x95) or (x85).

We also have (x89) and (x56), which seem to circumvent the "not be used to categorize" rule, but no such tags exist for fantasy, such as . Although, there is (x27).

Decades?

The decade tags have a definition that states they're for the 20th century (19xx), although it doesn't say whether it about works set in those decades, or works that were created in those decades. It seems primarily to be used for the latter. However, they're not used only for clearly-known time-frames. 13 questions are tagged both and , and 7 are tagged and , all with .

Decade tags have been burninated!

Movies, books, novels

We have and , which are both high-level genre tags, but used differently. To see how, we also have to bring in (which is plural, unlike the other two), which is defined as:

Use this tag only to differentiate the book from the movie or other media.

However, when you remove from the question, is used primarily to differentiate the work from the written media, rather than being the expected standard implied by the description, and is used more like says it should be. Keeping , and and are primarily used for that.

Oh, and there's (x66), which further entangles things.

This is still an issue that should be evaluated.

Hidden rules

I've found that there are rules regarding how to use specific tags hidden in the tag descriptions themselves, but I can't seem to find a meta that contains a list of all such tagging rules. For example, you can't go to one place to see how , and are supposed to be used. There's possibly more hidden rules, but I'm not going to go through every tag description to find them. In general, I don't think these description rules are followed. Instead, I think they're only enforced by those aware of the rules already.

Jedi, for example

Based on an earlier analysis of some tags, such as , I wanted to do an update.

Previously:

Out of over 1000 Star Wars questions, we currently have 66 questions tagged jedi, and that's one of the more popular tags. We have 887 questions tagged star-wars that contain the word "Jedi", but aren't tagged with jedi. This is an excellent example of why these sub-tags don't work.

Now:

Out of 2541 questions, we currently have 170 with . We have 598 questions without but contain the word "Jedi" in the title or body, 126 of which have "Jedi" in the title.

Discussion

Much of the meta discussions I found regarding some of the genre and media tags happened 4 years ago, when the site was young and smaller. There's merit behind the discussions, and I think they were valid and worked at that scale, but I'm not sure our general tagging policies now reflect how people actually use the tags. Here are some example questions:

Should we do away with the media tags (books, novel, TV, movies, etc.)?

  • This purged several media tags that weren't used very much at that time, but left others. By doing these purges, and actively removing attempts at recreating these tags that were previously purged, we can't get accurate numbers on how often people are trying to use them, which seems to be part of our criteria for whether or not we keep them.

How should we handle tag hierarchies?

  • This discussion is more recent, but has an interesting look at how some Work-specific tags are used, such as , and how Author tags are used, as well. The Star Wars stats have changed, so I decided to update them above.

https://scifi.meta.stackexchange.com/questions/606/horror-and-other-genre-tags

  • Discusses the creation of , which was poorly received but now has beat out our main genres!

Do we need a sci-fi and a fantasy tag? Why don't we have 'science-fiction' and 'fantasy' (or similar) tags?

  • These clarify why we have instead of . However, it doesn't address that people still feel the need to use the tags, and have co-opted the -genre tags to do the work for them when necessary. I think the and policies may need re-evaluation in light of the use of Probing questions on the stack. Probing questions make up a high number of our questions, and genre classifications can be helpful for them.
  • Sorry that it's gigantic. I can break it up into smaller bits if deemed necessary. – user31178 Jan 4 '16 at 21:08
  • 3
    I shudder to think how much spare time you must have... :-) – Rand al'Thor Jan 4 '16 at 21:17
  • @randal'thor Yeah, this was a lot more work than I expected, but business is slow. – user31178 Jan 4 '16 at 21:18
  • Although, to be fair, a lot of this was just done in Excel, and it took so long because my laptop is being weird and freezing up at odd times (I think the video driver is out of whack). All of the tag lists were generated from a simple query and a concatenate() formula. – user31178 Jan 4 '16 at 21:24
  • @creationedge ah, good old excel. – AncientSwordRage Jan 4 '16 at 22:56
  • A masterful analysis but lacking a central thesis. – Valorum Jan 5 '16 at 11:27
  • @Richard A thesis isn't my aim, here. Just the data with a little insight. – user31178 Jan 5 '16 at 14:47
  • No, but you must have generated some thoughts of your own while you were composing this essay. I'd be interested to know what those thoughts were. – Valorum Jan 5 '16 at 15:05
  • @Richard Yup, I've got some thoughts. I'm trying to be as objective in this thread as possible, though. I've got a couple more things I want to look at, and then I'll cook up more of an action plan or suggestions. – user31178 Jan 5 '16 at 15:46
  • This is why we should be able to award bounties on meta. – SQB Jan 6 '16 at 8:34
6

Categorization

Please feel free to discuss or disagree with how I did so, but I think my approach was consistent and unbiased. These are the categories I used in my other answer, but they were taking up too much space, there.

Franchise

Total: 14901

Probing

Total: 7729

Media

Total: 4501

Series

Total: 4268

Topic

Total: 4128

Work-specific

Total: 2665

Publisher

Total: 1363

Author

Total: 1211

Work

Total: 1103

Genre

Total: 1091


Top Tags

Here's an analysis of the the top 16 tags, and the top 9 tags used with each of those tags.

I couldn't figure out how to build a single query to do this for me, but I made one that gave me the top 9 tags used with each tag. (I meant to do top 10, but forget to update it before I did all my Excel work. The 1st result in the query is always the tag in question.)

Then, I ran that query for each of the top 16 tags. I chose the top 16 because they fit nicely in one Excel window for me, originally. Also, a few of them are closely related to one another, so I wanted more variety than just the top 10.

To clarify, this list is the top tags, by count, used in questions also tagged with the given tag, which happens to be from the top 16 tags used on the site.

The % given in each bullet is the amount given divided by the parent tag total.

  1. with 5496 questions

I expected the Genre & Media tags to be in the top 9, here. The Topic tags can be compared to my other post to see that each of their top tags is actually , too.

  1. with 2809 questions

  2. with 2504 questions

As pointed out before, isn't used in every question with the word "Jedi" in it, so if that inconsistency can be ironed out we'd likely see some different numbers there. I expect the same thing can be found with similar tags for both Star Wars, Harry Potter, and our other most popular franchises.

  1. with 2201 questions

The top tag here, about The Next Generation, is actually one of the top tags on its own. There's significant overlap here, but it's not complete. Clarity might be needed for how we use sub-tags and other tag hierarchies.

  1. with 1386 questions

The top tag here, again, is one of the top tags on its own, being about J RR Tolkien. is one of the only geographic tags (for example, there's no ). And, the here represent 63.9% of all questions with that tag (108).

  1. with 1106 questions

Simply, should not be on this list. Tagging MCU questions as movies is redundant, at least as it's being used. There could be an argument for TV shows being part of the MCU, but the movies tag isn't really used to differentiate between the movie MCU and the TV MCU, so the argument is currently moot.

  1. with 947 questions

Interesting, to me, is that is essentially just a single episode of a season of a series. It was broadcast separately, and sold that way, because of how the show is produced and marketed. I'm not sure any other popular shows have so many questions about special episodes.

  1. with 940 questions

  2. with 862 questions

The high number of Marvel Comics questions also tagged as comics seems to be used to specify Marvel property limited to comics only, which makes sense. is actually a synonym, and it looks less redundant if you think of it that way.

  1. with 856 questions

Again pointing out the overlap of The Next Generation with the Franchise Star Trek in general.

  1. with 738 questions

Every single question has at least one more tag, which means we may need re-evaluate how author tags are used.

What is the policy for author, series/universe/work tags?
and
General tagging practices
and
How should we handle tag hierarchies?
have different ideas about tagging authors, given at different points in time. I'm not sure what the "proper" use is, but we've been pretty liberal with our most popular author tag.

  1. with 635 questions

This tag is supposed to be for specific the book version of a work that's also a movie. So I'm not sure what is doing on this list. Checking out those dual-tagged questions seems like there's been incorrect tagging, in some way.

Even without this tag's definition, it's easy to see that shouldn't be on this list. That's redundant, no matter what.

This tag is often used just as is, especially with tags also used by Probing questions.

I think people primarily use this tag as a straight-up synonym for , so based on it's usage we may want to decide whether how people are using the two actually reflect what was discussed previously:
Tags: Novel or Book?

  1. with 628 questions

  2. with 613 questions

I don't watch the show, but I know that GoT is the HBO-show, based on the book series A Song of Ice and Fire. Knowing that, tagging it as TV seems redundant. Literary analysis should be done for the novels, not the TV show, and then I'm not sure if the book questions actually belong. We made need some tag cleanup there, because of user confusion between the two named franchises.

This tag is a good example of how questions about the TV/movie version of a franchise are clearly different that the written versions. However, most of our other major franchises, if not all of them, don't have a separate tag. For instance, there's no to separate it from .

But, with certain works we are quick to create modified tags of the work name, to separate it from other parts of the franchise. A clear example exists with , which not only includes the new Star Trek title movie, but also Star Trek: Into Darkness and presumably the upcoming Star Trek Beyond.

  1. with 592 questions

Over half the questions here use the Publisher tags.

  1. with 590 questions

And now the book version of this franchise. It is tagged with books, which seems redundant, and TV, which seems incorrect.

3

Naming Conventions

For some new works that have the same title as an existing tag, we append the year the new work/series was created:

Mostly, we use other conventions.

Media/Franchise tags

Name of starring character

This is a little harder to illustrate, as many of our starring characters have eponymous titles. I think we have a tendency to name this way, without seeing if there's a better alternative, anyway. The most clear evidence is
(x73), which exists even though is both more clear and within the tag character limit.

Use of subtitles

  1. Sometimes we use the full title and subtitle:

    But, as you see here, these two examples are not consistent, and one uses an abbreviation.

  2. Sometimes we use just the full subtitle

  3. Sometimes we use a shortened subtitle, even if the character limit isn't an issue

    Notice the lack of "the" here, but it exists for The Clone Wars tag and others, and "a" exists in (x590) and others.

Naming franchises

We have a lot of different ways we name franchises of connected movies and TV shows.

The official name for the collective works:

Sometimes abbreviated:

Then we also have:

In general, we have very inconsistent approaches to tagging works in a way that helps differentiate between characters, franchises, and single works, where two or three of those things would have the same name. This forces us into using the meta tags, such as and , sometimes in the same question.

A clear naming policy should be considered, that address how we handle franchise names, titles, and characters. Those questions that aren't overwhelmingly beyond retagging (such as ) should possibly be looked at.

The policy should also address the convention, such as when we abbreviate, use subtitles, or collective work names we choose vs. official names. While the name of the tag itself doesn't necessarily matter, if the wiki points to just fine, having a clear style will help guide people making new tags, and finding those before a wiki exists.

The most effective way of addressing this would likely be a "going forward" approach, rather than a massive "retag everything that exists contrary to our new policy" deal. (Supposing we ever make guidelines about this.)

  • This would have made a great question in itself (one I'd like to answer), but it's an excellent analysis regardless. – AncientSwordRage Feb 28 '16 at 7:17
  • Very thorough analysis (again)! Looks like it might make this old question obsolete. – Rand al'Thor Feb 28 '16 at 14:13

You must log in to answer this question.