Re: [engelang] Xorban: Termsets - engelang messages

On Sun, Sep 23, 2012 at 9:58 AM, And Rosta <and.rosta@hidden.email> wrote:

I'm very short of time to contribute. Brief comments:

1. "The benefits would be: getting rid of the “ne’V” series and using “jV” as they were intended to be used; a considerable amount of nonsense productions removed from the language; termset productions that would be close to underlying logical form.
The drawbacks would be a more complex grammar, including a new part of speech."
Other things being equal, the simpler grammar is to be preferred, but only when other things are equal. Simplicity of grammar is not one of the prime goals; rather, usability is. In this particular instance, it looks like the pros outweigh the cons.

There is a need for something like "ne/he", but I think we'd all feel better about delegating a whole C for "he" if there were other needs for non-formula complements. At this point of time, this would be a C for a single word.

2. All connectives are essentially abbreviations of an extensionally-defined set defined by explicit listing of members in combination with quantification over members of the set (e.g. "or" = at least one (is true); "and" = each (is true)), and quantificational predicates don't form a naturally closed class. For termset coordination, what is required is a list of ordered n-tuples, plus quantificational predicates, plus method of expressing an open proposition (i.e. property/relation), plus predicate (ck-, iirc) relating the open proposition to what binds the unbound variables within the open proposition. The special bits of grammar required for this are:
* method of listing -- doesn't exist yet
* method of listing ordered n-tuples -- doesn't exist yet (we had a method for ordered pairs, but not n-tuples).
* method of expressing open proposition -- we have this, but I can't remember now how we did it; at any rate, it's the method for doing "who loves who", either something like "la fa xi smi xu smu prmiku" or maybe "xiku prmiku".

E.g. "Some-but-not-all of {<Alice, Alfie, anemones>, <Bertha, Bill, begonias>, <Chloe, Charlie, chrysanthemums>} are in-the-relationship who-gave-who-what"

I have a separate idea for handling properties waiting in the wings that you may or not like, but I don't want to dump too much on this list at once. By the way, I guess I should apologize for using the word "termset" in reference to "ne". There is nothing at all constraining "ne" to tuples of terms. In fact "ne" permits and is intended to permit things like:

la nnla ni ne la nxla na ne le ckte nclake
“The boy does and the girl doesn’t like the book.”

My fuzzy memory of Lojban grammar had mixed up termsets and bridi tails. The "ne" particle works something like a leftward pointing bridi tail, or should we say, modifier tail. It's a means of admitting left-branching in a language which is primarily right-branching. Maybe we can call these things "termlists". We could rewrite the "he" rules as:

modifier := term | termlist

termlist := “he” explicit-sentence? | term termlist | “j” V explicit-sentence? termlist termlist

term := unary-operator | binary-operator formula

In this version, I have scrapped every production that isn't needed in more than one place.

I notice that (hypothetically) if we permitted all "n-" operators to be introduced into the grammar syncategorematically, then we could give different n-words different syntaxes. Since we don't have a lot of n-words (only three official), that's one idea for the future that might be better than using a whole C on a singleton (or small) word class.

Ordinary connectives are worth having special abbreviations for. But are termset coordinations? Are they frequent enough to warrant having special lexical and grammatical devices for abbreviating them? I'm doubtful. But at any rate, the priority must be first to ensure the underlying structure is expressible.