Sorry for the terrible pun.

Anyway:

pterosaur implies reptile while pteranodon (to which pterodactyl is aliased) implies dinosaur.

Pteranodons are a subgroup of pterosaurs, so pteranodon should either be aliased to pterosaur or should imply it - depending on whether or not we find it worthwhile drawing a distinction between various types of pterosaurs.

The pteranodon -> dinosaur implication ought to be removed - pterosaur should be the tag implying either dinosaur or reptile.

Technically, pterosaurs are flying reptiles, and not dinosuars. The pteranodon -> dinosaur implication is technically inaccurate, while the pterosaur -> reptile implication is technically accurate.

However, which of the two we go for is nonetheless up for debate. We often choose to go with common usage over technical accuracy, and most people searching for dinosaur would probably not be disappointed to find pterosaurs in their search.

Another possibility is to have pterosaur imply BOTH dinosaur and reptile - I don't think people searching for reptile would be too put out finding pterosaurs in their searches, so that route helps with both accuracy and common usage, while requiring no cleanup of tags to get rid of something we've decided to be a bad alias.

I would think, alias all of the particular types up to the top, implicate dinosaur from it, and implicate reptile from dinosaur. Does the system chain implications automatically? Yes, right? No need to make a redundant implication off of a particular type of dinosaur.

notnobody said:
I would think, alias all of the particular types up to the top, implicate dinosaur from it, and implicate reptile from dinosaur. Does the system chain implications automatically? Yes, right? No need to make a redundant implication off of a particular type of dinosaur.

I'm not sure a general dinosaur -> reptile implication is called for, particularly when we consider the number of feathered dinosaurs that really don't look much like the sorts of reptiles that most people would be looking for when searching for reptile.

But that doesn't make it not true, right? I mean like, dog implies canine, for instance. I notice tanuki doesn't, but it ought to, even they don't necessarily look the part. Since we can't bend reality to match any individual user's expectations, the best we can do is represent it as closely as possible and let any true thing be as likely as possible to result in the desired results. Like if someone only wanted the subset of dinosaurs that look like most common lizards, they could do dinosaur -feathers or something. Way easier for someone to come in the door and make a guess based on objectively true things than for them to try and predict what other people might have thought.

notnobody said:
But that doesn't make it not true, right? I mean like, dog implies canine, for instance. I notice tanuki doesn't, but it ought to, even they don't necessarily look the part. Since we can't bend reality to match any individual user's expectations, the best we can do is represent it as closely as possible and let any true thing be as likely as possible to result in the desired results. Like if someone only wanted the subset of dinosaurs that look like most common lizards, they could do dinosaur -feathers or something. Way easier for someone to come in the door and make a guess based on objectively true things than for them to try and predict what other people might have thought.

I just went over this in another thread.

I hadn't read that one yet, but I would've been on a similar page there too. I do agree some particularly obscure implications might fall below a threshold; I think that's the first I ever heard that orcas were dolphins, and I'm not sure I even buy it. But dinosaurs being reptiles, I think that's pretty hard to argue isn't well known, even if it isn't mentioned all the time. Right on with fixing pterodon->dinosaur mistags, which would mean they should just also imply reptile themselves. But I still think implication trees should be used. We can't possibly know in every case what usage is common among any and every subset of people, but we can definitively know what's actually correct, and if correct is the only option, correct and common become the same thing for everyone. It's definitely a frustration I can relate to when your own usage is different from what's right, but it's way more annoying when your usage is correct, but correct isn't supported because someone else decided their own version was better. And with tree setups, it's simple enough to prune with a - if it really matters.

notnobody said:
I think that's the first I ever heard that orcas were dolphins, and I'm not sure I even buy it.

https://en.wikipedia.org/wiki/Killer_whale

>The killer whale or orca (Orcinus orca) is a toothed whale belonging to the oceanic dolphin family, of which it is the largest member.

Heh. Idea bought. Still like genjar's tree.

From the oceanic dolphin wiki link:

"Dolphins, alongside other cetaceans, belong to the clade Cetartiodactyla with even-toed ungulates, and their closest living relatives are the hippopotamuses"

furthering my agreement that not all links are worth adding. Might not complain about seeing flipper when I meant to see shamu, but I'd be pretty surprised if I saw peter potamus.

notnobody said:
I do agree some particularly obscure implications might fall below a threshold;

It is well known that humans are apes. Do you support a human -> ape implication?

It's less well known that apes are monkeys. Do you support an ape -> monkey implication?

Both are correct. Neither are common usage, but the former is well known all the same. How exactly do you distinguish "too obscure" from "this is correct so we should go with it"?

Personally, I think this is the key point: Both would impede searching by forcing users to constantly exclude humans with a '-' from their searches. Therefore both are bad implications. Common usage trumps taxonomic accuracy.

That's the key to my argument that you don't seem to have addressed.

Human to ape, yes, but I could see someone's argument that humans are a sufficiently special class to make an exception. Ape to monkey, I'm not really familiar with the topic, but some cursory reading makes it look like biologists don't seem to agree on it. I could see being swayed either way. I don't know that I've really worked out what I think would draw the line for too obscure, but I'd definitely be more inclined to deal with having to narrow things out when a net winds up being too wide than missing things I thought I would get when I search what I figured was an umbrella term. It's an easier problem to fix because it's visible. You see a lizard with feathers and don't want feathers, take away the feathers or ignore those results. But if you want to see all of the canines on the site, and all you get are six out of the thirty kinds we have images of, how do you know what you missed? You can't really find that out. Could be that the better route is to only do implications on types, not individual species? Trying to think about whether that would keep the same question going anyway, especially when individual ones already link back to the types, e.g. husky->dog->canine.

notnobody said:
But if you want to see all of the canines on the site, and all you get are six out of the thirty kinds we have images of, how do you know what you missed? You can't really find that out.

Umbrella tags aren't always called for though.

For instance. Birds are a subset of dinosaurs (specifically theropod dinosaurs). However, people searching for dinosaurs are almost certainly not looking for birds, and forcing them to exclude the term every single time is obnoxious and impairs usability - makes getting the results you want less intuitive and it takes up a tag slot for an exclusion, effectively reducing the number of search terms available.

In other cases, incorrect umbrella terms are useful: again, implying pterosaur and icthyosaur to dinosaur is generally useful, even if inaccurate.

There is not a catch-all solution for these things. And even if there were, "taxonomic accuracy" is not that solution.

Yeah, I don't think strict taxonomic accuracy is necessarily useful. Not everybody's a biologist. But I feel like there's got to be some solution we could come up with that can serve the purpose without making things up, even if it can't actually be validly used all the time. One guideline that pops to mind would be disqualifying any link where you have to use the word "technically." "Technically, birds are dinosaurs." Eh. That link's proposed, and seems like a clear descendant relationship for sure, but if you go with

noun: dinosaur; plural noun: dinosaurs
1. a fossil reptile of the Mesozoic era, often reaching an enormous size.

You could break the link really quickly. Like, just to keep harping on it, a dog isn't just technically a canine. It's true and happily pretty much everyone knows it. An orca's a dolphin, but, ehhhh,...just technically? I mean, that tends toward your common usage strategy, but maybe that should serve more as a disqualifier for an attempt at building trees, more than the opposite where, say, people usually misrepresent hyenas as canines and since that's common we go ahead and wrongly include them. Then sometimes you might miss some things in a search you thought was an umbrella, but there's not much room to complain because it's like if you wanted to find a book when you actually searched for a dvd or something.

I mean that's pretty much just a roundabout way of saying "common usage", except with the reasonable qualifier of "not when it's too terribly wrong", which basically turns it into what I said earlier of "tend towards common usage, but there's no catch-all strategy, so we have to take things on a case-by-case basis sometimes".

Maybe so. So what's the takeaway? Seems like we get more utility out of trying to get more results into a search, even if tagging of some of the potential result images lags behind. More tags always means more results on arbitrary searches in general. And more truth to reality seems to tend toward more tags, based on the examples, even if that ends up being overkill. So what's the strategy? I'd think the way forward is to look for any possible links between anything and only disqualify them when it's disputed or such trivia fodder that it would actively fuck up people's attempts to get decent results. At least, from the other end of it, is there ever a time when it would make real sense to actually build a lie into the tags? I wouldn't think so. I'm more okay with a missed search that creates a need to learn than a missed search that creates a need to un-learn.

My takeaway is that the best way to go about this is to ask "is this what people are looking for when they search for [x]?". I don't know - that's how I go about things anyway.

In the case of dinosaurs and reptiles, I doubt it. In the case of pterosaurs and reptiles, maybe, but maybe not - we can argue that point certainly and I could see it going both ways.

Either way fixing the issue of pterosaur and pteranodon being disconnected tags is the first priority.