tag categorization

Is there any interest in further tag categorization? It might be useful to have categories such as clothing, hair, body, pose, setting, item, format, etc. to differentiate one type of tag from another. If this is something we might be interested in, what categories would be important / useful.

I've gone through the taglist and categorized all the tags with more than 50 posts, at least where there there is a fairly obvious categorization. Unfortunately that's only 8% of the tags or so, since over 75% of the tags have less than 10 posts. Anyway if anyone's interested, I could post my categorizations somewhere.

Also I was thinking as I went through all these tags cleaning stuff up as I went. 37% of all tags only have one post, and a large number of them are typos or variants of a proper tag. Might it be a good idea to warn a user when they are creating a new tag, so that if it's supposed to be something pre-existing they can fix it?

Additional tag types are a nice idea, but adding complexity to a system makes it less likely for people to use it. Characters, copyrights, and artist represent a good set of minimal types because at a glance they'll tell you who drew the picture, what show it belongs to, and who the character is. Typing a tag as a hair type, while nice for organizational purposes, isn't as of much use to the typical user.

I agree, and it's probably a good idea to hide the non-essential categorizations from the user. Having them still accessible via API, or via search could be useful though. Say you wanted to search for pictures where a character's hair color wasn't tagged, or something like that.

It might also be useful to anyone building an application that would enumerate the tags, so they could be sorted by what they described. It would also be easier to tell accidental tags from intentional ones. If a pre-existing, but uncommon tag was categorized, it's more likely to be legitimate, if it's uncategorized, it has a better chance of being an accidental addition.

Right now 70% of tags are "Other" which really isn't all that useful. If we filled those categorizations with some sort of metadata, even if we didn't show it, it could be a useful addition.

Rather than tag categories, relationships types would be more descriptive. That way we could distinguish between artist X drawing a picture and artist X appearing in artist Y's omake, or between a picture of series Z and a picture of series W parodying series Z. But these are edge cases. I like the simplicity of the current approach.

The reason I'm saying categories is a good idea is because they are already implemented. Relationships would entirely change the way the system works and (probably) make it too inefficient to run as a web-service with many users.

Although it would be nice to be able to see the semantic relationship between hair.red and hair.blonde for example or even (ayanami_rei.hair.blue to attach tags to their referrents), that's pretty impractical. The ability to attach a number (category) to a tag is already built into the system. So you could still sort of see the relationship between tags: red_hair:6 ~ blonde_hair:6 in the same way that ayanami_rei:4 ~ ikari_gendou:4 .

Shinjidude said: Is there any interest in further tag categorization? It might be useful to have categories such as clothing, hair, body, pose, setting, item, format, etc. to differentiate one type of tag from another. If this is something we might be interested in, what categories would be important / useful.

This was discussed when we first hammered out the idea of tagtypes. Initially there was talk of extending them further, but we decided against it. There's a upper limit where getting more complicated doesn't really improve functionality. I think extending tagtypes much more crosses that line, and making EVERY tag typed is really asking for trouble.

jxh2154 said:
This was discussed when we first hammered out the idea of tagtypes. Initially there was talk of extending them further, but we decided against it. There's a upper limit where getting more complicated doesn't really improve functionality. I think extending tagtypes much more crosses that line, and making EVERY tag typed is really asking for trouble.

Hmm, I was around when the artist, ambiguious, copy, & character tags were first added, and don't remember any discussion on extending it. Perhaps it was after Danbooru started hosting it's own forums, because I didn't frequent the new forum too much. In any case if that is the consensus I won't push.

I don't see how it could cause any trouble though, being that you could choose to not show the tags to the user (it would be invisible), and the mechanism is already in place so it wouldn't change the complexity of the system itself.

Also, EVERY tag would be infeasible to categorize, primarily due to the fact that there are a HUGE number of almost unused tags, most of which are probably redundant with something else, erroneous, or irrelevant. For the most commonly used tags, I still hold that there would be a use to further categorization.