[lojban] (to all coders) what's missing: semantic analysis

[purpleposeidon <purpleposeidon@gmail.com>]

A Dandy Gentleman Wrote:
> I've got somewhere a list of what
> selma'o I think are atomic. Will share later.

I can't find the list, so I'll make it up.

  I, IJE, SE, tanru, TUhE, TUhU, atomic sumti

tanru: would have JE broken up, and also BE ripped out. CO might be
untangled, but would be kind of useless because tanru are pretty vague
anyways. But it might be nice to determine that "le slaka lerfu" and
"le lerfu co slaka" could refer to the same thing.

atomic sumti: zo ratni, da, zoi gy. Canada! .gy., and lots of other
stuff for things like joi-masses.




In Timo's examples:
> mi pinxe ki'u lo du'u mi taske <-> mi pinxe i ki'u bo mi taske

the decomposition na gerna. (And also, I don't really know how to use
bo.)  What I'd go for is {mi pinxe .i krinu go'ixi-1 .i <a selbri for
du'u (I don't know if one exists. A pseudo-gismu might be needed)>
go'ixi-1 .i mi taske}


> lo gerku e mi e do cu klama lo ricfoi <-> lo gerku cu klama lo ricfoi i
> je mi klama lo ricfoi i je do klama lo ricfoi
{
  .i daxi0 ricfoi
  .itu'u daxi1 gerku .i daxi1 klama .i daxi0 se klama tu'e
  .ijetu'u mi klama .i daxi0 se klama tu'e
  .ijetu'u do klama .i daxi0 se klama tu'e
}
(do and friends actually wouldn't be used that way, they'd be a
daxiny. to current listener or whoever -- that's how I'd do it,
anyways)



So, actually implementing it.
.ysai Some preprocessing *may* be required. This might involve some
portion of daxiny.-ing atomic sumti, untangling FA and SE, ... . I'm
not certain how much of this is necessary.
For doing the Action step, I'm pretty sure the best thing to do is to
have an Annotated Grammar. I'll attempt to make some pseudo-code out
of a few complicated rules. The <> apply to the previous construct.

selbri-4 =
    (selbri-5 [
        (joik-jek selbri-5
        )<create a new semantoid, using joik-jek as its JE, and
selbri-5 as its tanru>
      | (joik [stag] KE # selbri-3 /KEhE#/
        )<create a new semantoid, using joik as its JE, and selbri-3
as its tanru. [stag] would be filed away under some misc bit. >
    ] ...
    )<selbri-3 would have a list of semantoids. Any semantoids created
would be added to that list.>

bridi-tail =
    bridi-tail-1 <Using the inherited semantoid-cluster, and a new
list of terms that would be applied to that semantoid>
    [gihek [stag] KE # bridi-tail /KEhE#/ tail-terms]
    <Create a new semantoid-cluster. Any terms created prior to selbri
in bridi-tail-1 would be included. [stag] goes to misc. Create a
semantoid-cluster, give it to bridi-tail and tail-terms.>

sumti-tail =
      ([sumti-6 [relative-clauses]] sumti-tail-1
      )<create a new semantoid-cluster. Pass it to sumti-tail-1.
Create a daxiny., assign it to the new SC's x1. Give the daxiny. to
the parent. If sumti-6: Create a semantoid-cluster. The selbri is
"<association>". Put sumti-tail-1 into the x1, and sumti-6 into the
x2. And if relative-clauses: Define sumti-6 as ke'a and let it run
wild>
    | (relative-clauses sumti-tail-1
      )<create semantoid-cluster. Set the hypothetical x1 to ke'a (or
make up a daxiny for ke'a and latter merge it with the x1). Run
relative-clauses. Run sumti-tail-1>


A semantoid-cluster seems to be something that has a bunch of terms,
and a tanru; whereas a vanilla semantoid would have only a single term
and a SE tanru. They'll have to be written with much more care than
I've put into them. For example, sumti-6 must certainly work when
accessed via sumti-5, but it also must work from sumti-tail. I was
hoping that some simple markup could go inside the <>'s, but it will
probably have to be code.


ALSO: I have a question for jbocevni and jbosampla: BNF or PEG? Camxes
is PEG, and everybody loves camxes. But it is in Java, and I don't
know Java (and I think timos doesn't either; I dunno about other
Lojbanists). However, ba'e zirsam .o'a uses BNF, so I am much more
familiar with the BNF.

mu'o mi'e .timos. si .djeims.

-- 
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.