@And: response re termsets in the works.
On Wed, Sep 19, 2012 at 8:46 PM, Jorge Llambías
<jjllambias@hidden.email> wrote:
> Here is a way to implement selpa'i's proposal easily in the current
> grammar. Call "p-" the "afterthought restriction operator". It would be a
> binary operator that would look like a binder, but instead of binding a new
> variable, it would conjoin another restriction to the existing restriction
> of that variable.
That seems more palatable than adding a new selma'o. If we can be sure
it doesn't cause problems.
> se mlte sa grka pe xkra vskake
> "Some cat, some dog, a black cat I mean, [that dog] sees [that black cat].
>
> This would be logically equivalent to:
> se je mlte xkre sa grka vskake.
What happens when there's no prior restriction:
se pe xkre mlte bjre
Would that be the same as "je"?
I hadn't intended for "pe" to be used in the very restriction that it was intended to "conjoin" its own restriction with. That would be infinitely recursive. Therefore, your example would only make sense in something like
_se1 sme1_ se2 pe xkre mlte bjre
which would mean effectively
se1 sme1 se2 pe1 xkre1 mlte1 bjre2
which in turn means:
se2 sme2 bjre2
And what happens when you can't move it back:
re mlte sa grka pe jrsake xkre
Every cat, some dog, that the dog chases the cat, is black.
I took a fresh look at this, and determined that if "pe" were an "afterthought restriction conjoiner", then it could to a certain extent be formalized as a "smart" logical operator that knows when to be "je" or "jana" by looking back at the operation (s or r) that bound "e". Here are some examples (in each, a,b,c all mean the same)
"Some cat, a black one, is happy"
s1a) se mlte pe xkre glke
s1b) se mlte je xkre glke
s1c) se je xkre mlte glke
"Some cat, some dog chased by, runs"
s2a) se mlte sa grka pe jrsake bjre
s2b) se mlte sa grka je jrsake bjre
s2c) se je sa grka jrsake mlte bjre
"Every cat, if black, is happy"
r1a) re mlte pe xkre glke
r1b) re mlte jana xkre glke
r1c) re je xkre mlte glke
"Every cat, every dog, if chased by one, runs"
r2a) re mlte ra grka pe jrsake bjre
r2b) re mlte ra grka jana jrsake bjre
r2c) re je sa grka jrsake mlte bjre
Mixed cases:
"Some cat, every dog chased by, runs"
s3a) se mlte ra grka pe jrsake bjre
s3b) se mlte ra grka je jrsake bjre
s3c) se je ra grka jrsake mlte bjre
"Every cat, some dog, if chased by that, runs"
r3a) re mlte sa grka pe jrsake bjre
r3b) re mlte sa grka jana jrsake bjre
r3c) re je ra grka jrsake mlte bjre
So to a *certain* extent, in response to selpa'i's question, we already have an "afterthought restriction conjoiner"; I just hadn't thought of it that way. It's called "je/jana" i.e. "jana" with "r-" and "je" otherwise. I say "a certain extent", because everything is fine up until the first "ju" (what I sometimes call "the main predication"). After that, ju-semantics kick in and you're no longer in restriction-land. I still plan to study "p-" for that reason, but I doubt I'd ask for a whole consonant. Maybe a jV'Vk- or something.
(As a minor note, did you ever consider assigning an primitive operator for "jana"? I know technically we don't need it, but it'd feel nice to have "->AB" along side "v~AB".)
One last thing: "p-" did have one ability that "je/jana" cannot handle, which I think could probably be handled with termsets, but I will save that for another email.
That looks donkeyish:
re frmre sa xsla pe pnseka drxeka
Every farmer, some donkey, that the farmer owns the donkey, beats it.
They would be donkey cases if the main predication had an "a". The complexity here comes from wanting to conjoin a previous restriction with a new restriction containing a variable free in the previous one, but that can be fixed by binding the that variable inside the new restriction, which I indirectly showed.
co ma'a mke