[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
On Tue, Aug 21, 2012 at 11:31 PM, Mike S. <maikxlx@gmail.com> wrote:
>
> I hope no one minds me branching this off into a new thread. The Loccan3 thread is scattered, and I think (what I hope no one minds me tentatively dubbing) Xorban deserves to be focused on.
The only thing with that name is that it's not a valid name in the
language itself. Before coming up with a proper name though it would
be better to have a more clear idea of how to form predicates. I'm
thinking names will be predicates like almost all other words of the
language. Here's the grammar we have so far again for reference:
sentence := predicate | operator sentence
operator := unary-operator | binary-operator
binary-operator:= SV('V)* sentence
unary-operator := NV('V)*
predicate := CCC*V('V)*(KV('V)*)*
V := a | e | i | o | u
C:= b | c | d | f | g | j | k | l | m | n | p | r | s | t | v | w | x | y | z
S:= l | m | r | s | ...
N:= n | ...
K:= k | f | ...
> Okay, what you're basically saying here is that "l-" doesn't encode specificity nor does it quantify, in which case I'd say it doesn't mean anything, nor does it do anything other than perhaps occasionally introduce a variable that one is just going to repeat later in the predicate. Why not just drop "l-" and let predicates introduce their own throw-away variables when the need arises? What's wrong with:
>
> - nnla rzge nxli dndakeki
> = A/the boy gives a/the rose to a/the girl.
That parses as four sentences: (nnla) (rzge) (nxli) (dndakeki)
The obvious rule is that disconnected sentences are assumed to be
connected with je (order doesn't matter here) so that would be
equivalent to:
je nnla je rzge je nxli dndakeki
or equivalently:
je je je nnla rzge nxli dndakeki
But this would only work at the highest level. It wouldn't work for
example if we want to apply an operator to that whole thing, then we
need the explicit "je"s.
We also could have some rule about how unbound variables are bound,
and binding with "l- sm-", where "sm" is a predicate that is true of
anything is an obvious choice.
> - nnla nxli dndakeki
> = A/the boy gives [something] to a/the girl.
>
> - sa nnla rzge ri nxli dndakeki
> = Some boy gives a/the rose to every girl.
Here it's getting odd though. That's two sentences: (sa nnla rzge) (ri
nxli dndakeki), and the "a" of the second sentence is outside the
scope of "sa nnla". To keep it within scope we need:
sa nnla je rzge ri nxli dndakeki
which is now a single sentence with an unbound "e".
> - sa nnla ri nxli dndakeki
> = Some boy gives [something] to every girl.
>
> If a variable has to be introduced early for some scope-related reason, use a generic predicate like perhaps {zst}.
As a minor side issue, I would want to keep the charged predicate "...
exists" well distinct from the empty predicate true of anything "...
is something".
> If we can safely drop "l-" for constant, maybe we can recycle it to allow specificity to be marked optionally, which is something I often find useful (I may be biased due to being English-L1 though):
>
> - la nnla rzge nxli dndakeki
> = The[=a specific one, it should be clear which] boy gives a/the rose to a/the girl.
That parses as three sentences:
(la nnla rzge) (nxli) (dndakeki)
again with "a" outside the scope of "la nnla".
>> > Are there situations in which variables like {a} "stick"
>> > and become essentially anaphora, and if so when?
>>
>> I don't know. We could have some of the reserved variables behave this
>> way, but that would mean you have to plan in advance which argumants
>> you would want to keep around.
>
> It seems natural to suggest a series of V'V-space e.g. {o'V} that automatically declare anaphora, at least now for experimental purposes.
Maybe we could just say that an unbound variable already used before
is assumed to be bound by "l-" and with the same restriction it was
last bound. So for example:
na sa xrja vfla ... na se nlceka
It is not the case that for some x which is a pig, x flies ... It is
not the case that for some y, y is a wing of x (=pigs).
No pig flies. They have no wings.
Here I note I've already used "nlc" both for "x1 likes x2" and for "x1
is a wing of x2". We will need some vocabulary construction if we
don't want this kind of homonymy.
> At any rate, Xorban will eventually have to allow adverbial expressions, including whose with arguments. Morneau's system is just one possible rationalization of a universal need.
The first step is to see how they are dealt with in predicate logic.
If we use events, then for example:
"The horse runs fast."
The x Horse(x): Ey Event(y): Runs(x,y) & Fast(y)
la xrma se fwe je bjrafe stre
(With fwe: "e is an event") Or perhaps even:
The x Horse(x): The y Something(y): Runs(x,y) & Fast(y)
la xrma je bjrafe stre
with an implicit "le sme" binding for unbound "e".
>> > It's a promising start. I hope that you will continue to develop it.
>>
>> Only if there are more questions. :)
>
> You have these. Other questions are ready to be loaded on trucks and sent your way. :)
Keep them coming, and tell me what you think of the suggestions above.