Danbooru

nuke gen_x_pokemon tags

Posted under Tags

nonamethanks said:

I wasn't exaggerating.

gen_2_pokemon: 9.7k posts
gen_2_pokemon -gen_1_pokemon -gen_3_pokemon -gen_4_pokemon -gen_5_pokemon -gen_6_pokemon -gen_7_pokemon -gen_8_pokemon: 4.1k posts

For gen 3 it's 11k vs 6k.

That's thousands and thousands of posts with padding tags.

Yes, that is exaggerating.
See it this way: You always receive an image where a gen 2 Pokémon is present, in all of its variations. That's kinda the purpose of the tags, wouldn't you agree? You may already expecting lots of crossovers with other generations, as well as some everyone pics.

Because, let's think about the alternative: You actually search for or with a specific Pokémon, like Sneasel. You would never use the gen_2_pokemon tag with it, because it's redundant: Sneasel is from gen 2.
However, if you're interested you may be looking for an instance where Sneasel is present with any gen 5 Pokémon. Then you can use the gen tags. And if you want a specific set, you can still search for it, like Sneasel+Charizard.

Gen X Pokémon is just a very broad and unspecific tag. That makes it very flexible to use. You might say it's too flexible because of all the everyone tags in it, seemingly watering down the tag.

See, I would have agreed with you there in theory.

But then I went to look at the analytics and I couldn't find even a single search combination of gen_*_pokemon with any other pokemon. Most search combinations with these tags are weird searches done by a single guy like gen_4_pokemon black_gloves or gen_8_pokemon beer, or gardening searches like gen_*_pokemon -pokemon order:score status:any. I don't see evidence that these tags are actually useful to anyone.

nonamethanks said:

I couldn't find even a single search combination of gen_*_pokemon with any other pokemon.

In this case, I assume the analytics are broken because I have done that kind of search before sometimes. Plus there could be future searches like Provence said before.

Maybe the name "gen_1_pokemon" can be considered unclear? If it helps, maybe it could be renamed to "generation_1_pokemon" or "pokemon_(generation_1)".

nonamethanks said:

All our other group/team tags require a focus or presence of multiple characters belonging to it in the tag, why should pokemon work different? Imagine if we tagged every single idolmaster or love live girl with their idol group.

You're saying Generation 1 Pokémon is a group/team tag?

I'm thinking it's like tagging a cat a cat. Or tagging a Pokémon RGBY character pokemon_rgby. Gen 1 Pokémon just means the character was introduced in the Gen 1. It's almost like a copyright tag.

In fact, Ash Ketchum is a Gen 1 character as well (but we don't have a tag for humans of each generation). Kanto is a Gen 1 region. Poké Ball is a Gen 1 item. Tackle is a Gen 1 move. It's simply the generation where something was introduced.

I definitely don't think it's the same as a tag for the band members. Going by the logic of idol groups, gen_1_pokemon would basically be a tag specifically for all the 151 Pokémon from that gen together, right? (post #3028663)

Danielx21 said:

In this case, I assume the analytics are broken because I have done that kind of search before sometimes. Plus there could be future searches like Provence said before.

The analytics only show the last week of data. However I checked it at multiple times during the month and it was always the same irrelevant searches.

Danielx21 said:

I'm thinking it's like tagging a cat a cat.

No, this would be like tagging every single cat, dog, bear, bird, snake and other spine-having animal with vertebrate.

Updated

I've made comments about this on Discord, but I'll reiterate my stance for posterity.

The gen_*_pokemon tags were created as a compromise to solve the issue of copyright tags getting used with Pokemon, which I agree was an issue since they should only need the base copyright tag of pokemon. However, my feeling is that we can continue our stance of not using specific copyright tags with Pokemon, to avoid the myriad of searching issues that would result, and still nuke gen_*_pokemon tags; it's not one or the other.

Therefore, the issue instead becomes: how useful are the gen_*_pokemon tags for searching? I have personally never used these tags in my searches, so it's unsurprising to hear other users generally aren't either. I also agree with the point that using these tags in isolation is unlikely to result in the desired posts due to the prevalence of cross-generational images.

I think tags functioning as "gen_*_pokemon_focus" would be much more useful. At that point, though, we'd be better off starting from scratch, which makes nuking the existing ones a non-issue.

Updated

nonamethanks said:

See, I would have agreed with you there in theory.

But then I went to look at the analytics and I couldn't find even a single search combination of gen_*_pokemon with any other pokemon. Most search combinations with these tags are weird searches done by a single guy like gen_4_pokemon black_gloves or gen_8_pokemon beer, or gardening searches like gen_*_pokemon -pokemon order:score status:any. I don't see evidence that these tags are actually useful to anyone.

Tbh. it's always better to keep the possibility for the future there, even if it's not searched at all now.

nonamethanks said:

For consistency I checked today again and these are the searches for these tags this week. Even more abysmal than two weeks ago.

So looks like these tags got used 311 times in a week. This would mean approximately 16.172 times in a year (311 × 52).

So they are definitely used. Good to know.

I don't think you understand what kind of search traffic we get. Any random tag can have that amount of traffic due to the sheer size of the site. That gen 2 pokemon, a 10k posts tag, has the same traffic as pink wristband (a tag close to 100 times smaller) should tell you how useless it is.

But sure, fine, let's keep them around. Looking forward to see how we deal with them once gen 9 comes out and only one user will be mass adding the tag while everyone else ignores it.

blindVigil said:

For comparison, Genshin Impact gets more than 10k searches a day. That's 33 times what these pokemon gen tags get in a week, approximately 233 times the activity, while only being 3 times the size. 16k searches a year is practically nothing.

Just for the sake of statistics, my link is IPs searching for it, not searches themselves.

I'm using NewRelic to calculate the data, I don't know why the popular searches page got those numbers but they're *a lot* smaller and have been for a long time. I haven't checked the reportbooru code but I think they might be sampled or something.

Looking at the actual analytics, searches containing "genshin_impact" (so including character tags with qualifiers) in the last 24 hours alone were 353k across 30k different IPs. 1/3rd of a million searches for that copyright alone IN A DAY.

By comparison we get on average between 300k and 400k searches per hour. In the last hour there were around 500k unique post views. That's the kind of traffic we get.

Updated

I absolutely understand the purpose of these tags existing - they serve a generic-copyright tag role, intending to reduce the tag padding that comes from folks being idiotic and tagging the game copyright rather than just the overarching Pokemon tag. In effect, it isn't there to be searched for (even if there some folks searching them), but to act as a solution to a meta-tagging problem. However, as had been brought up at the start, this isn't exactly an issue the Touhou copyrights suffer.

The Touhou copyright wiki clearly states:

NOTE: Don't tag individual Touhou games unless the game is directly referenced by the post. All of the characters in a picture having debuted in the same game is not enough to count. The full cast of a game (with or without the protagonists) appearing will usually count, so long as no other characters are present.

You could easily include such a clause in help:pokemon tags when clarifying on tagging media; in fact, it already does!

A post depicting a Pokémon species is not necessarily a visible reference to any game in particular.

For instance, adding pokemon_swsh to a Wooloo post (post #3528342) would be a mistake, because Wooloo is expected to appear in other media as well.

As a longer example, post #1015491 is a Moltres. It should be tagged gen_1_pokemon. Technically Moltres first appeared in RGBY, but it should not be tagged pokemon_rgby. Moltres can be a reference to any of multiple games and media. It can be captured in multiple games such as FRLG, XD, HGSS, XY, and USUM, not to mention traded to other games. It also appears in multiple manga, the anime, and at least one movie. That's why, instead of using more than half a dozen media tags for Moltres or other Pokémon (which would be technically correct, if impractical), we use just the generation tags as explained above.

Using the Touhou approach would remove the need for that last paragraph - just broaden the previous sentences and you're good to go! Of course, even the Touhou approach has had its own vagueness issues, as per topic #19184, but it is certainly something that works. In fact, if these tags were removed, then you'd effectively got such an approach immediately, thanks to the tag gardening that had been done to make these tags in the first place. If one is diligent enough to tag these, then they're diligent enough to garden without them. And if there are repeat offenders, just slap them with the above-linked help page.

So, while I understand keeping them, given everything, it does unfortunately make more sense to get rid of them and just concentrate one's efforts in gardening directly.

Updated

nonamethanks said:

But sure, fine, let's keep them around. Looking forward to see how we deal with them once gen 9 comes out and only one user will be mass adding the tag while everyone else ignores it.

One user? Looks like Provence populated most Pokémon gen tags initially, back in 2018.

Anyway, I assume some people will once again try and tag new gen 9 Pokémon with the game tag, and that will kinda make sense until the same Pokémon species appear in the anime too, in the manga and other later games. This kind of tagging happened in gens 7 and 8, at least.

If that happens with gen 9 too, I guess I'll have to remove the game tags and add gen_9_pokemon. That's important clean up work. Yeah, I've not been much active on Danbooru but I guess I can do that anyway.

Here's a list of the top 500 *pokemon* searches this week:

Popular Pokemon characters get hundreds or thousands of unique searches per week. These get less than a hundred. This is very low considering their size, and how many searches other tags of a similar size get.

The problem with these tags is that they have a very low value-to-weight ratio. They appear in the tag list on every Pokemon (creature) post, sometimes multiple times if there are multiple Pokemon, yet very few people use them. I don't think they're worth the weight.

1 2