User:Quercus solaris/sets

Preamble

 * Sketches of filtered sets: hypernymous ontologic nodes (noun phrases or open compound nouns) and their contrast sets of cohyponyms, plus related terms, see also (cf), and so on
 * To an extent, the existing population of Wikipedian categories is on about this same set of meta-themes
 * The potential population is vast (almost unimaginably so); the gaps remain legion (quite understandably so); nonetheless, humans haven't done any better anywhere else, excepting certain KRR efforts, some of which are behind walls of various types and heights (some walls are higher than others; some walls are thicker than others; some walls are pricier than others)
 * Boosters keep threatening that generative AI is about three seconds away from solving this problem more thoroughly than humans themselves have managed (or can manage) to solve it; from some viewpoints, it seems like they're correct (about the imminence); but my gut keeps nagging that a truly whole-assed version of an AI solution is probably not yet as close as they imagine. Just imagine if my gut is wrong about that. Strangely, either way (yes or no), a lot of humans are in for a lot of surprise, in one way or another.
 * The root of the rub is that the chatbot layer itself is fundamentally stupid and insane, no matter how well it collocates the bits and pieces that make up its word salad. We're told that "sometime soon", other AI layers (or AI-ish layers) that are more about KRR will be layered over the insanely stupid layer to provide noninsane/nonstupid ontology. The reason I have so much doubt about the claims is a combination of (1) "if that's so straightforward then why hasn't it been done yet, or at least fairly started" plus (2) the queasy realization that apparently half or more of humans don't even understand the difference anyway — they seem to genuinely have no clue about, and possibly no ability to conceive mentally at all, the qualitative epistemic difference between "human liars who might be bullshitting or scamming" and "chatbots spewing confabulations at orders of magnitude higher volume." That difference is roughly the difference between a mere pepper mill and Why the Sea is Salt, but they apparently have no mental capacity for conceiving that difference? I don't know, I'm just reacting to the evidence that I've seen so far. If we're lucky, I've misunderstood what I've seen so far, and later the mechanism of misunderstanding will become clearer.
 * One thing about the potential setups and schemas is that a queryable database is a relevant model, and one would like to ask and refine iteratively. And that's a party trick that generative AI is getting pretty good at (iterative prompt engineering, where the machine remembers and understands the preceding context/state). You know those things with the kajiggers on them? Yes. How many kinds of those are the ones with purple kajiggers specifically? Here is a list. Are you sure you aren't just [the functional equivalent of] high on LSD? Yes. Did you remember to include mauve kajiggers too? Here is a revised list where the purple category is understood to subsume the mauve category. Which ones of these are commercially available in my region? Here is a list. How many of those are under 200 bucks? Here is a list. Remember that guy who was working on developing a yellow kajigger? Yes. His name is X Mc-Xerson. What's he up to these days? Here is a list of his most recent publications. Do you think we'll ever have blue kajiggers? The quest for blue kajiggers is ongoing. Here is a list of recent publications.
 * Handwave etc
 * It occurs to me that senseid is obviously one key piece of the overall puzzle. A sense is often such a combinatorial as is discussed above (a combinatorial hypernymic node), and it can have sensewise relations (e.g, syn, cot, hyper, hypo, mer, hol, rel, also; the last two must be at their own section, but they can have sense labels, which ties back to senseid anyway, so still good to go). At moments this might almost seem like the whole picture, except for one obvious limitation: you can't have any senses or senseids if you don't have a headword to put them under, and you can't have a headword unless it meets CFI. This is an interesting constraint of the medium, like painting without a full color palette perhaps (I'll have to keep sniffing at that analogy, in a nonolfactory way). From here (given this limitation), go to the other ideas about combinatorial hypernymic nodes elsewhere herein.
 * Related instance: utterance or vocalization as hypernymic node (e.g., Thesaurus:utter); a cardinal first level of cohyponymous nodes below that is melodious or harmonious utterance or vocalization (e.g., Thesaurus:sing, Thesaurus:hum) versus nonmelodious or nonharmonious utterance or vocalization (e.g., Thesaurus:speak, Thesaurus:talk, Thesaurus:shout, Thesaurus:murmur)
 * sonic emission is the hypernymic grandparent above utterance or vocalization (i.e., generation from nonhuman or human source); cardinal contrast sets at that level include (1) nonhuman versus human; (2) nonhuman-and-nonhumanoid versus human-or-humanoid; (3) animalian-including-human versus nonanimalian (with either of those subsuming humanoid, depending on a value judgment or viewpoint); nonsentient versus sentient (but you'd have to choose where you'd put pseudosentient in that dichotomization, which is analogous to nonhuman-but-humanoid vs nonhuman-and-nonhumanoid)
 * Some other spaces that are not irrelevant to this exploration include Categorization and Appendix:Roget's thesaurus classification.

What-if daydreams
These are things that perhaps will never happen at Wiktionary, but imagine for a moment if they happened at some Wiktionarylike place? Not so much a fork as a spork or tritensil·ʷᵖ. Not all forks are contentious; one of the points of CC licenses is cordial remixability. And (speaking of utensils) as long as we're mixing cordials and cocktails, don't forget your swizzler.


 * Imagine a dictionary with a 'Not to be confused with' subhead. Such a thing could sometimes be useful to EFL learners.
 * What if any adjective could have a subsection for "cardinal things that are often describable with this adjective"? (Collapsible of course; collapsed by default per a setting; could not must) (e.g., spoonlike > ladle)
 * What if any adjective could have a subsection for "other [semi-cardinal] things that are sometimes describable with this adjective"?
 * Obviously there are cutoff values for relevance; which leaves …
 * What if any adjective could have a subsection for "anything else that is potentially describable with this adjective"?
 * … which admittedly is a dog's breakfast not to be examined by humans unless they are digging for something, but …
 * what if you had a fleet of a million AI bots that could aggregate such things in the background?
 * What if there were a dedicated method of recognition for cardinal coinstantiations that are highly relevant Venn overlaps albeit not synonymous or hyponymous, but are recognizably cohyponymous beneath some hypernymous node of noun phrase that is substantively akin to a Wikipedian category?
 * One relevant semantic distinction that Thesaurus already formally recognizes is the distinction between hyponym and instance. It is but one small step further to formally recognize cardinal coinstantiations that are highly relevant Venn overlaps albeit not synonymous or hyponymous, but are recognizably cohyponymous beneath some hypernymous node of noun phrase that is substantively akin to a Wikipedian category (e.g., interrogatee > arrestee, detainee, suspect). If Wiktionary does not care to do so, there is room in the world for something else Wiktionarylike to do so. In fact Wikipedian categories already almost are that thing (even now), but there is a certain tightness missing that a more Wiktionarylike thing could backfill.
 * Note that earlier thesauri have lumped these things into the same lump as syn or ana (aka near-syn). One of the themes that was an original thrust of Thesaurus was to do one better (or several better) than they had done, making due use of the digital medium to its full advantage. This is just a case in point.
 * What if any word's coordinate terms section could be branched into subsections with a label for each subcategory, such as sense labels, combinatorial labels such as "things that are a type of X and have traits Y and/or Z", and so on, with each of those divisible into subsubsections for "cardinal or archetypal subset", "logically comprehensive subset" [of less use to humans but not zero], and handwave etc?
 * What if you built a dictionary that was dead serious about simply and explicitly labeling the vertical polysemy subclass of polysemy (i.e., the autohyponymy subclass)? (The difference is [total] Venn subsumption versus Venn overlap.) Traditional lexicographic methods include the distinction between a sense and a subsense, which is nicely labeled in most competent dictionaries (e.g., "1a" or "1.1", subordination indention), and this method goes a long way (i.e., most of the way) toward handling the whole aspect; still, one function that its form, by convention, cannot do, is illustrate this relationship without putting the (auto-)hypernymous sense (the wider one) first (i.e., giving it first-listed position). But sometimes the (auto-)hyponymous sense (the preciser one) wants to be first-listed (i.e., when it is the primary sense, as ranked by importance, being most common in use, or both, where importance can include a component whereby precision lends weighting). An example is cow versus cow, whereas some cows are cower than others, according to the silly way that we humans are neurally predisposed to talk. (In fairness, it's not the predisposition per se that is silly, but rather the prevailing degree of incapability to override it.) There are potential ways to mark/flag/tag this aspect in a way that is not list-position-dependent and would not be difficult to do, explain, or understand (for either writer or reader). I say this granting sight and touch as senses with which to convey the labeling; not so much hearing (that one's not at all impossible, but I don't see easiness or intuitiveness for it). The sighted ones include shading conventions and icons. The touch ones (considered especially, but not at all exclusively, for nonsighted readers) might include several shaping conventions that suggest themselves readily enough. The latter class would be quite useful in AR forms, for both sighted and nonsighted users (you don't have to be sightless to appreciate eyelessness, in a non-eyeless way). Enough for now. (If you want a flying car, you have to build it yourself, as the rest of these clowns, even the KRR postgrad ones, tend to talk big and perform small.)

Cells > blood cells

 * See blood cell § Hyponyms
 * Note: There is also the distinction (to be handled better eventually) of (1) practical subset RBC+PLT+WBC versus (2) "everything else" hematopoietically speaking (e.g., e.g., e.g.); return to this; again this comes down to the human mind's "practical" and "archetypal" groupings versus the set of all etically relevant groupings

Drug classes > currently clinically relevant members

 * See handwave etc
 * See w:Category:Drugs and its subcategories

Mechanics > diesel mechanics > pet peeves > customer-originated

 * "It bugs me when they do that", where they = X
 * Go

Mechanics > diesel mechanics > pet peeves > engineer-originated

 * "It bugs me when they do that", where they = Y
 * Go

Quantities > practical > abstract > stuff

 * See Thesaurus:lot