ralju brivla deep structure

From Lojban
Revision as of 18:45, 21 October 2024 by Korvo (talk | contribs) (→‎Gismu Deep Structure: Part 2: list counterexamples. Lots of links here, folks; if you don't grok, read up before commenting.)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
lo si'o lo ralju brivla cu banzu = The idea that main verbs are enough

The Ralju Brivla Deep Structure hypothesis (RBDS hypothesis) is the hypothesis that the set of main verbs (mostly gismu) constitute a complete vector space for the expression of all meaning. In other words: that the Lojban main verbs set is sufficient to express all possible predicates, without separate lujvo, without tanru, and (here's the controversial bit) without prepositions. It presumes that all of these can be expressed with sentences (inordinately complex and ugly sentences, true) involving just the set of main verbs and cmavo (other than prepositions.)

A more narrow hypothesis is called "Gismu Deep Structure hypothesis" (GDS hypothesis) which states that not main verbs but only gismu (thus a subset of main verbs) are enough. As explained below, GDS is false.

Gismu Deep Structure

The original formulation of gismu deep structure is from la nitcion:

GDS (nitcion 1)
All lujvo can be expressed as tanru.
GDS (nitcion 2)
All tanru and lujvo can be expressed as bridi containing only gismu and cmavo.

In (1), la nitcion also says, "And if we admit that not all concepts can be expressed with gismu and cmavo, … well, I'd worry." A hardliner might therefore propose a bolder hypothesis like the one at the top of this page:

GDS (vector-space)
The set of baseline gismu generate a vector space covering semantic space.
GDS (completeness)
The set of baseline gismu can express all semantic concepts.

As originally given by la nitcion, the hypotheses are more like guiding rules for how to define Lojban vlaste than concrete statements about (baseline) Lojban; they aren't falsifiable. In contrast, the hypotheses about semantics are completely backwards. In wikipedia:model theory, the models of a logical system always have a predictable structure: there is an initial model, also called a standard model or term model, whose semantics are precisely those which are logically required by the syntax; and many non-standard models, which definitionally contain concepts that cannot be expressed via standard syntax. A standard introduction to this way of thinking is studying wikipedia:non-standard model of arithmetic.

The pedantic type theorists in the audience should note that the initiality conjecture has been proven for Martin-Löf's dependent type theory, which includes Lojban's second-order logic as a special case, so Lojban's categories of models will also have a term model as initial model.

Robin Lee Powell said in 2024 [1] that the falsity of GDS "[is] not a point of contention."

Counterexamples

The following concepts can't be semantically addressed via baseline supplication, either because supplication was insufficient or because the concept challenges the ambit of second-order logic and must be approached axiomatically.

{xlane}
The proposed fourth tense of Lojban; its exclusion from baseline coined the phrase "Whorfian mind-lock", later included in the wikipedia:Jargon File slang dictionary
entropy
a generic concept relating knowledge of a system to observations of a system; one of many recent scientific concepts missing from baseline
SI units
there are gismu for individual units but no baseline mechanism for forming tensor products of units and thus no Abelian group of SI units
algebra
a system for manipulating syntax with equational reasoning; these aren't in the baseline and can't be built without internalized syntax-manipulation tools which also aren't in baseline
algebraic theories
systems governing algebras, particularly wikipedia:Lawvere theory; individual systems are addressed at abstract Algebra, but there is no way to specify additional other systems
category
like {cmima} but for arrowhood in a category rather than elementhood in a set; must be given as a primitive concept because there are ontological distinctions between Lojban's assumed wikipedia:Zermelo–Fraenkel set theory and the wikipedia:Tarski–Grothendieck set theory required for category theory

Support and criticism

  • It is not known what the minimum set of main verbs is to be. It is nevertheless agreed that most verbs in this set are gismu.
  • Some Lojbanists (e.g. Adam and Nick) thought most if not all prepositions in Lojban can be paraphrased as bridi: see pe necessary for sumti plus (BAI-type modifier) Gotcha.
  • seljvajvo suggest ways in which lujvo (and by extension tanru) can be expressed as explicit combinations of gismu (be lo, be lo nu, je).
  • most fu'ivla can be mostly (messily and verbosely) described in terms of pure Lojban, or at the very worst, foisted onto cmene.

History

The hypothesis is probably the inception of hardlinerism; it proposes that much of Lojban grammar and semantics can be reduced to a simpler core. It was first formulated explicitly (unsurprisingly) by Nick Nicholas, in his lujvo papers[1][2], which formulated seljvajvo. However, it is implicit in the very existence of the cmavo ta'u, which certainly predates him.

"Deep Structure" is Nick's allusion to Generative Grammar, particularly in its earlier, explicitly transformational form. The implication is specifically that other facets of Lojban grammar can be explicitly derived from this "Deep Structure", which is semantically primary.

Nick:

This reveals my bias towards Generative Semantics

Discussion

  • nitcion:
    OK, you can start yelling now :-)
    • Jay:
      Why yell? Seems like its probably right. I don't think its a good reason to do anything special, though. You could express the same things without anaphora, but you're certainly not going to stop using KOhA anytime soon. :)
      • nitcion:
        You're right, of course. This isn't so much a reason to do anything different, as a underlying motivation for why certain things are already being done.
      • xod:
        If 1350 primitives (times however many places) are "enough to express all meaning", why is that number special? Suppose there are 10 000 different sumti places. Why is 10 000 enough primitives but 10 too few? Such a statement means that the actual number of primitives in reality is some number between 10 and 10 000! Who is prepared to make that claim?
      • Adam:
        Of course the actual gismu list is not minimal. There are many gismu which can paraphrased in terms of other gismu. You can't really tell whether any particular set of gismu is minimal, because it's always possible that you overlooked some paraphrase, or that you cut some malglico corner in reducing the list. Also, there's always a large number of concepts which need fu'ivla. Still, the gismu list appears sufficient for covering the vast majority of semantic space.
        • xod:
          By introducing the overlapping of certain gismu you are adjusting my extremely hypothetical figure of 10 000. My point holds just the same, whether 10 000 is adjusted down to 1000 or up to 1 000 000 actual unique gismu places. I assume that, as you introduce fu'ivla, and say "vast majority of semantic space" instead of "all of semantic space", you are disagreeing with the notion, proven false above, that "gismu constitute a complete vector space for the expression of all meaning".
        • nitcion:
          The theory, which may or may not be trivially false, is not as important as the ideology:
          1. it is not a futile or unilluminating endeavour to find GDS's for concepts (hence, lujvo, rather than fu'ivla, whenever possible);
          2. it is possible to disambiguate concepts entirely within Lojban (hence, disambiguation of candidate tanru senses through GDS, and appeal to such disambiguation in seljvajvo)
          3. it is possible to understand Lojban grammar and semantics by using a common core of Lojban (i.e. the relation between fi'e and finti is not accidental or irrelevant, but on the contrary formalisable.)
          Your attitude to this, as with hardlinerism in general, is "it is not possible for all conceivable cases, therefore it is futile." My attitude is "it is possible for just enough cases to make it a worthwhile pursuit." And indeed, that pursuit motivates my interest in Lojban. Wittgenstein passed over in silence whereof he could not speak; that doesn't mean he thought he had been wasting his time. (Actually, maybe he did; he was weird like that. I don't have to. :-) You know, maybe one day I will do the Tractatus in Lojban.) Similarly, Gödel's Incompleteness Theorem proved that Bertrand Russell's attempt to formulate an internally consistent theory of mathematics in the Principia Mathematica could not cover everything, because paradoxes are still possible. It did not prove that mathematical logic is pointless.
          I'm reminded of the Neogrammarians here. The neogrammarians, with typical Teutonic rigour, argued that all linguistic change was regular. They were wrong factually. But they were very right methodologically: you only throw your hands up in the air, and say that an instance of language change is irregular, when you've made damn sure you've exhausted all regular possibilities. Otherwise, you're not doing historical linguistics.
          • Adam:
            Well said, and I completely agree, but I don't think you pay attention to it enough. I mean, is "skepre" the best you can do for "scientist"? .i'o xamgu se cusku
            • nitcion:
              I never said I was unwilling to follow constructive criticism, Adam :-) In fact, inasmuch as I was unable at the time to come up with anything better, I strongly urge you to. (And as you know, xorxes pointed out about me that I metatalk hardliner, but use Lojban like a naturalist. Maybe I'm being realistic despite myself; maybe, on the other hand, I'm being lazy. It's hard for me to tell. Like I say, I want nothing more than to discuss this, and what alternatives there might be.)
    • Maybe skeplicre in the sense used in Epictetus.
      • xod:
        My attitude is rather consistent with Gödel's; I have outlined a proof that shows that no number of semantic primitives is enough to cover all meaning. That was the position to which I am responding, wasn't it? As for pointlessness, I don't know anything about it. The more interesting question is: are the 10 000 primitives a subset of those commonly used in English? (If so, we are playing with a Simplified English.)
  • jezrax:
    This is an interesting philosophical discussion, but it seems to me that it doesn't quite grapple with the issue. Let me try out a few points:
    • People can grasp meanings that are not expressible in natural language.
      • The expression on somebody's face can be described, but the description doesn't convey its full meaning.
    • Every natural language can express meanings that are unavailable in other languages.
      • Try to fully explain the German idea of Gemütlichkeit in another language. You can get close, but you can never nail it.
      • See this poem of mine. The phrase "works and days" in this poem has a literal meaning, makes a literary reference, and makes two internal (punning) references. One can't translate such things in a way that keeps the full meaning.
    • Lojban's gismu are unquestionably capable of expressing everything that's pragmatically necessary.
  • .i iuro'e uonai Let me try to follow out my line of reasoning.
    • The hypothesis that "gismu constitute a complete vector space for the expression of all meaning" has two components. First, the gismu must cover "meaning space", whatever that is.
      • That's easy. broda covers meaning space all by itself.
    • Second, it must be possible to combine gismu to indicate any meaning in meaning space.
      • That's hard. In fact, I think that my examples above show that it is false.
    • This begs the question: What is the meaning space? Is it the set of all ideas that some person could, in principle, grasp? Is it the set of ideas that some one person could in principle communicate to some given other person? Is it the set of ideas that can in principle be communicated by a spoken language? Does it rely on the set of ideas that everyone can understand, that somebody can understand, that some person could in principle understand though in fact nobody living could? Et cetera.
      • I rather think that meaning space itself is ill-defined.
  • So I conclude that the Gismu Deep Structure hypothesis, "gismu constitute a complete vector space for the expression of all meaning," is by itself neither true nor false; it is ill-defined (in the mathematician's sense). One needs to decide what it means before one can figure out whether it's true. (zo'o of course, you can say that about anything in philosophy!)
  • But I also think that the GDS hypothesis is false under reasonable assumptions about the meaning space. English speakers know when to say "a" and when to say "the" and what they both mean, but I defy anyone to write a complete explanation, of any length, in any language, of what exactly the two words mean. Russian speakers know when to use perfective aspect and when to use imperfective and what they both mean, but again, I defy anyone to make a complete explanation. A language community may come up with a word for any concept they can agree on; I don't see a guarantee that that concept must be even comprehensible to outsiders, much less exactly expressible in an outside language.
  • A more general question is whether Lojban has the same level of expressiveness as a natural language. To me it seems to. It loses expressiveness by having a small vocabulary (translate "we dawdled on the scree"), but it gains by precision and flexibility.
  • la gleki:
    • IMO, gismu deep structure means you can write a complete definition of any word (or particular word when applied to it) in pure Lojban using only baseline gismu and cmavo.

Links

References