[X] My Assistant

A decent metric of the overall effectiveness of a tag is the ratio of upvoted galleries with the tag to total (including downvoted) galleries with the tag. (example, f:moral degeneration has 376 upvoted out of 1141 total, coming out to 33% accuracy overall. That's pretty bad)

I look at these stats a lot and got fed up with doing it manually every time, so here's something I made for fun. It's a big list of all tags on the wiki (male/female/misc/language only), and shows their overall accuracy for each namespace.

[reasoningtheory.net] https://reasoningtheory.net/accuracy



Click the top buttons to sort. This could be useful doing cleanup, or for finding trends, or for looking at which tags and types of tags to be very very careful about. Etc. I thought I'd post it because maybe someone other than me will find it interesting.

Some things I'm seeing... costume tags are generally pretty good, 85%-95% much of the time. Activity tags are mediocre; in f:, excluding f:solo action and one-image tags, the highest one is the unusual f:ttf threesome, at only 88%. (most of them range from ~85% to 60%. I guess the presence rule is still too arcane for some to grasp) Infidelity tags are a mess, around 70%. A lot of the "weird" or unusual tags are in the 50s and 60s. Most of the tags in the 40s and below are animal-related, I guess due to confusion with 1-image bestiality, furries vs animals, and namespace issues. (is the namespace the one who is X or the one being sexed by the X?)

Some caveats.

• Some of the tags' inaccuracy is due to old namespacing (like f:crossdressing) rather than tagger failure. The same can be said for tags whose definitions have changed.
• A few of the tags have horrible accuracy simply due to having been blocked/slaved in one of the namespaces
• The tags that have few galleries have a high accuracy variance. So the 100% accuracy of m:doll joints doesn't mean much since it has only 4 galleries.
• The true accuracy of nearly every tag is certainly lower than what is listed; there still exist hundreds of thousands of galleries with bad tags that haven't gotten cleanup attention yet.

This post has been edited by Superlatanium: Jan 9 2017, 13:53

I'm actually rather surprised by how inaccurate a lot of these are. As you say, it's really not all that hard to get accurate results - I'm only a somewhat-active user, not super in-tune to the goings on, but I've made just under 20k tag votes just off the cuff in my time here and ended up with a Started Accuracy of 94.1 % and a Voted Accuracy of 98 %. Literally all you need to do is have a bit of common sense and the ability to go "does this fetish show up on more than just one or two pages?"

It actually kinda weirds me out, seeing as we have systems in place (mod power) to make sure the people who don't reasonably know what they're doing don't have a huge effect; best guess I can make that doesn't involve incompetency is that people might upload without knowing a lot about tagging mechanics, tag everything they can see with the first thing that works for them ("I'll try putting 'fox' in, there's a foxgirl!), and then quickly get their low-power tags downvoted by vets.

I've always liked these metrics, thanks for puttin' em together. Also, if you want to check your own accuracy, remember there's the tool at >
ehentaihip.com/tools.php?act=taglist&uid=
to do so with - just put your user ID (at the end of the URL if you click into your userpage) at the end of that address to check, though you're only allowed to do it for yourself. Check if you're part of the problem or the solution!

Also, if you want to check your own accuracy, remember there's the tool at >
ehentaihip.com/tools.php?act=taglist&uid=
to do so with - just put your user ID (at the end of the URL if you click into your userpage) at the end of that address to check, though you're only allowed to do it for yourself. Check if you're part of the problem or the solution!

Also, if you want to figure out exactly what you've done wrong (which everyone has, at some time or another), use Tag Check to save your upvoted tags and check to see which ones have gotten downvoted.

This is just my opinion, but I think a good goal for those who understand English, want to be accurate (and aren't afraid to look at the wiki whenever they aren't entirely sure) is something around 98% accuracy.

It might take only 0.5 seconds for someone who wants points to upvote a tag, but it can take 10 seconds - 2 minutes for someone doing cleanup to be sure that a tag is invalid and downvote it (depending on how long the gallery is and how easy the tag is to identify). That's why I think even 95% accuracy is troublingly low, especially for those with a lot of mod power and high tag volume.

It actually kinda weirds me out, seeing as we have systems in place (mod power) to make sure the people who don't reasonably know what they're doing don't have a huge effect; best guess I can make that doesn't involve incompetency is that people might upload without knowing a lot about tagging mechanics, tag everything they can see with the first thing that works for them ("I'll try putting 'fox' in, there's a foxgirl!), and then quickly get their low-power tags downvoted by vets.

An issue is that nowadays, everyone and their dog has 20+ mod power. Well, maybe not quite, but many do, and power is no guarantee that they know what they're doing. The other issue is that tag points are greatly incentivized, and for many, they only really care about tag accuracy to the point of "don't be so inaccurate that you get tagging privileges revoked." It's not that people can't figure out how to be more accurate, it's that there's next to no incentive for it, so many don't bother. At least, that's my guess.

thanks for the tag score link. though you mentioned an ENGLISH speaker should clear 98% accuracy, when most galleries are Japanese, and unless one is REAL fluent, they may tag quite wrong

my personal rate is "started: 80% of 635" "voted: 78% of 149" and 991 tags overall

i apologize if this post is not useful

my personal rate is "started: 80% of 635" "voted: 78% of 149" and 991 tags overall

i apologize if this post is not useful

It's useful insofar as it tells us you are not particularly good at tagging.

Please read tag definitions and be sure you know how they apply before adding them.

Tags: 22061 (136 recent) Started Accuracy = 96% of 784 Voted Accuracy = 99% of 20879

I'm definitely not the most active of users, but even so, this is what I've gotten. Hah. The thing is, from what I can tell, the inaccuracy when it comes to the misuse of various tags comes not just from the misunderstanding about the tagging rules... but also because quite a few people don't seem to understand what a bunch of tags actually mean.

Take the example of double penetration. The wiki describes it as:

"Description: Two orifices (mouth, vagina, or ass) being penetrated simultaneously. Very likely to invoke the group tag."

Yet if you EVER add the tag for situations where a character gets repeatedly spit-roasted, but doesn't get the "classic" ass+pussy double-penetration, you end up with a bunch of down-votes, even if there's actually a LOT of double-penetration in the gallery.

I doubt this is the only example where the user base seems, on average, unable to understand what a tag actually means, much less when it's appropriate to use it.

And this was my random two cents before I most likely go back to lurking.

Interesting. Does this update live or on some delay? Or is this just a snapshot?
Does this account for removed or expunged galleries? Since we aren't tagging those.

Can this be set up examine average user tagging accuracies (and then cross reference)?
It would be interesting to see any correlations with mistakes between starting/voting scores vs mistagging.
Or even mistagging trends in high scoring taggers.

Does this account for removed or expunged galleries? Since we aren't tagging those.

We aren't?

We aren't?

Not after they are removed or expunged at least.

Tagging expunged galleries is not necessary but there are no actual rules against doing so.

Expunged galleries may come back, e.g. plain bad expunge or expunge war. If the gallery contains fjordable content better have it tagged; when dxed the gallery will appear on some far numbered page and may lose scrutiny.

i.e. in several cases it is a good idea to tag expunged galleries anyway.

Fair enough, it does seem like it could be easily exploited though to just throw tags all over galleries that may well not be searched.

Tagging expunged galleries is not necessary but there are no actual rules against doing so.

Expunged galleries may come back, e.g. plain bad expunge or expunge war. If the gallery contains fjordable content better have it tagged; when dxed the gallery will appear on some far numbered page and may lose scrutiny.

i.e. in several cases it is a good idea to tag expunged galleries anyway.

That's good to hear. I don't particularly go out of my way to tag expunged galleries, but when I land on them for whatever reason, I occasionally check it and tag accordingly.

Fair enough, it does seem like it could be easily exploited though to just throw tags all over galleries that may well not be searched.

Would someone specifically do this just on expunged galleries? I would have thought those that want to be an arse and just tag whatever would do it on all galleries. Guess you can't know what every random knob that doesn't care to be a proper part of the community thinks about though.