Re: Another Sketch: Palno

On Tue, Aug 26, 2008 at 2:06 AM, Alex Fink <000024@...> wrote:
> On Mon, 25 Aug 2008 23:04:40 -0400, Logan Kearsley <chronosurfer@...>
> wrote:
>
>>On Mon, Aug 25, 2008 at 10:04 PM, Alex Fink <000024@...> wrote:
>>> Given that clauses can be zero-converted to arguments just as atoms can, is
>>> there really a distinction between atoms and predicates?  Why not analyse
>>> atoms as 0-ary predicates?
>>
>>Mm... because I didn't think of that? It probably is a valid analysis,
>>if you allow single atoms to be sentences.
>
> Indeed.  And this is what I did.  One of the design goals of my language was
> to have one open word class, so I didn't want any distinctions based on
> arity.  Another, for that matter, was strict compositionality.
Another possible arrangement (though one that doesn't help with
further reducing the number of classes) would be to consider
noun-like-things to have infinite arity, verb-like-things to have
restricted arity, and modifier-like-things to have zero arity.
Noun-like-things (including complete clauses) would take
modifier-like-things as arguments and return a reference to a more
specific instance of the noun-like-thing.

>>I think it's useful to make the distinction between things that
>>represent individual noun-like ideas, and things that describe
>>relations between other simpler ideas, though.
>
> I'm not sure about useful, but if nothing else it's probably something well
> embedded in the human linguistic capacity.  I was trying to go without such,
> anyay.
One of my design goals here is to see how easily comprehensible I can
make it with the minimum of disruption to the basic underlying logic,
so some concessions to things that are embedded in the human
linguistic capacity are good.

> Of course what a noun-like idea is a fuzzy boundary.  {Adjectives} in langs
> without a class of adjectives are treated nominally in some and verbally in
> others (and we've taken opposite approaches on this, it seems -- I made
> adjectives zero-ary and consequently had great amounts of the appositive
> adposition).
That was more first approach, but I did not like the deep nesting
structures and/or extended conjunctions necessary to apply multiple
descriptors. And saves on a morpheme and makes the structure simpler
if descriptors are treated as 1-ary predicates.

>>> I had an engelang whose grammar was of a similar sort, and eventually I got
>>> really tired of the inelegance and inflexibility of different fixed argument
>>> sets and return values, as it were, of various types of word.  [...]
>>> So klugey.
>>
>>Can you elaborate on what's klugey about it? Or is it just an
>>un-analyzable matter of aesthetic preference?
>
> In most part it was, if not violation of the one open word class goal, then
> at least not having smoothed the distinctions between its subclasses out as
> I might have wanted to.  (So aesthetic, but at a higher level.)
>
> Something more akin to the system in Rikchik (which I didn't know of at the
> time) would've satisfied me better.
The Rikchik system does seem similar to what I'm using; I have 3
arities where Rikchik has 8 or 9, and I use case marking instead of
theta-role marking, but the basic logic is I think the same.

>>> So.  Do you have a way of rendering relative clauses, or anything like them?
>>
>>Yes. There are two ways of doing it, one which conforms to the postfix
>>parse tree, and one which temporarily sets up its own parsing
>>environment.
>>Sample sentence: I like people who eat apples.
>>Method one uses a determiner to pick out the relevant case of the
>>argument that's being modified:
>>I ((people these) apples-ACC eat)-ACC like.
>
> Ah, okay.  So you only provide one level of scope?  the relative anaphor
> (er, cataphor) points to the nearest clause containing it, always?
Yes.

>>Method two uses a relative pronoun to indicate that you should pause
>>the current clause and start parsing a new one, and sets of the
>>relative clause with commas:
>>I people-ACC, who apples-ACC eat, like.
>
> This seems to have potential for ambiguity, if one's not fastidious about
> the positions of the commas (and maybe even if one is, with a more
> complicated example -- I haven't convinced myself either way).  Consider a
> sentence of form
>  a who a.ACC p a.ACC p a.ACC p
> where each a is an atom, and each p a predicate taking nom and acc
> arguments.  Does the relative clause finish after the first p, or the second?
Or after the third, leaving an incomplete sentence. You can't tell
without the commas. They are absolutely required. It would be nice to
have a way around that, but I haven't found one yet.
Keeping up the mathematical/computational analogy, my justification
for this structure is that a comma + relative pronoun is like a
function call that temporarily creates a new stack frame to parse the
next bit of the sentence in, and the closing comma is like a return
statement that restores the original stack frame, but with the top
element altered.

>>I'm pretty sure every case can be handled with the first form, but the
>>second form I think is more generally human-comprehensible.
>
> I wouldn't make that leap.  There are plenty of natlangs with internally
> headed relative clauses, which is your first strategy.  Standard Average
> European speaker comprehensible, maybe.
I should re-phrase- neither structure is basically better or worse
than the other, but I think it's easier to untangle complex or
multiple nested instances of the second than it is to work out all of
the relations in more complex and multiply nested instances of the
first.
That could just be me; maybe some other people do have less difficulty
following the parse tree for the first form.

> Anyway, slightly sharper example of the "schizophrenia" I was talking about.
>  Suppose you have a sentence of form ... hm, what's not a too pragmatically
> strained example?
>  arrow [hand me.ACC of].ACC through partial
> Is this "an arrow is partially through my hand", taking the adjective at the
> end as a clause adverbial?  Or is it ["the arrow through my hand is partial"]
> , taking the internal clause as nominal in force?  Supposing it's just
> one of these, how would you say the other?
That would be "an arrow is partially through my hand."
"The arrow through my hand is partial" could be said in several ways:
"[[arrow this] [hand me.ACC of].ACC through] partial"
"arrow, REL [hand me.ACC of].ACC through, partial"
Or, as "there is a partial arrow through my hand":
"arrow partial [hand me.ACC of].ACC through"

Generally, when applying a predicate to a predicate argument, it's
supposed to apply to the relationship (the composite object of all of
the arguments and how they are related to each other) defined by the
predicate argument. So, when modifying a through-relation with
'partial', you have to get 'partially through'. There are situations
where pragmatics says that it only makes sense for the new relation to
apply to a sub-part of the predicate argument, and that gets you a
de-facto semantic distinction between adverb-like predicates and
adjective-like predicates, but it's a very fuzzy one.

-l.

Replies