Print

Print


On 9 August 2015 at 03:27, And Rosta <[log in to unmask]> wrote:
> On 9 Aug 2015 01:52, "Logan Kearsley" <[log in to unmask]> wrote:
>>
>> On 8 August 2015 at 08:58, And Rosta <[log in to unmask]> wrote:
>
>> > But I certainly wouldn't agree that it isn't possible for there to be a
>> > syntax such that there is a rational consensus that it has just one PoS.
>> > Setting aside special quotative devices, my Livagian has one PoS, and
> this
>> > is not especially impressive or surprising. The one PoS corresponds to
> the
>> > logical category of predicates. The cobinding of arguments, standardly
>> > notated with variables, is expressed by inflections. Trad logic's
>> > quantifiers, connectives and constants are also treated as predicates.
>> > There are no syntactic subcategories of predicate. So just one PoS.
>>
>>
>> Palno is similar (some versions of it, anyway- others have two or
>> three, due to splitting out atoms and conjunctions from predicates).
>> I don't know how well this would apply to Livagian, but my Devil's
>> Advocate position on Palno is that, even in the version designed with
>> only the one part of speech for predicates, can reasonably be argued
>> to have 4 parts of speech by an outside observer, based on the
>> different syntactic behavior of predicates with 0, 1, 2, and 3 arity.
>
> For Livagian, the outside observer would see only one PoS. I take it that
> the essence of a PoS is that it allows you to attribute more than one
> property to (the members of) the same set of lexical items.
>
>> There are, however, definitely more than one kind of internal
>> syntactic node in all versions of Palno.
>
> Not for Livagian. It is completely left-branching and every node (terminal
> and non-) is a predicate. The mother node is a complex predicate formed of
> its daughters.

I am getting the impression that Livagian must have some similarity in
function to J. S. Jones's indexing languages. Or possibly Davin, the
"phonology on top of set theory" language.

Either way, I suspect that major difference between Livagian and Palno
that results in Livagian only needing a single lexical category is
that Palno associates arguments with sub-expressions, and thus the
morphology on each sub-expression is required to express whether it is
representing a particular relativized argument of itself, or the
proposition, which results in two kinds of syntax nodes. Whereas
referential arguments are not identified with any one specific phrase
in Livagian, instead being implicitly described by the inflections
distributed across multiple monocategorial predicates. Does that sound
right?

>> I also can't speak as to how well Livagian avoids the linguistic
>> equivalent of the Turing Tarpit, but as I have noted before I find
>> basic monocategorial Palno *incredibly* difficult to actually use
>> fluently. Experience with other conlangs suggests to me that this is
>> to some extent inherent to the language, not purely a result of my own
>> unfamiliarity with that kind of language. So I am still waiting to be
>> convinced that an unambiguous monocategorial language is not merely
>> *possible*, but that one can be made *ergonomic* and *practical*.
>
> My intuition is that Livagian cannot be processed as a human language.

Well, then, I guess the search continues!

> I don't think that monocategoriality or any other sort of excessive
> simplicity is the main problem. The main problem is how to keep track of
> which argument places across the sentence share the same value. I know of
> just two sorts of solution to that. One is Livagian's, which is to maintain
> an abstract list of argument places in the sentence so far, and encode with
> inflections interactions -- list-merging operations -- between the
> arguments of the rightmost word in a phrase and the arguments of the
> sentence so far. The other solution is every other loglang I know of's,
> including standard notation, which is to give each argument a unique (but
> recyclable) name. This latter solution is even more unergonomic than the
> former.

The Palno solution is not quite the same as that standard loglang
solution. As stated above, it identifies every argument with a
relativized phrase that projects the referent of that argument as its
semantic head. If one argument needs to be used in more than one
position, where one of the positions is not an ancestor or descendent
node of the other, it uses pronominal predicates; but, while that
maintains total syntactic disambiguity, there is room for ambiguity in
pronoun binding. I don't worry about that too much because I consider
it roughly on the same level as lexical ambiguity and polysemy, which
I don't any loglang can completely eliminate- you'd need and infinite
number of basic roots for every possible way of slicing up subsets of
the world!

This exchange has, however, brought me to a couple of
Obvious-In-Retrospect realizations:

First, Patrick's language can trivial be transformed into a truly
monocategorial-at-the-lexical-level by introducing some regular
inflectional morphology for distinguishing preposition-like and
noun-like words, rather than encoding the distinction in the roots
themselves (although that does retain a distinction between two types
of internal syntactic nodes).

Second, there is a logical equivalence between isolating bicategorial
languages and inflecting monocategorial languages; and one language
could be analyzed as either due to the ambiguity of word boundaries
and the fuzzy line between morphology and syntax. In any case, a
usable language must encode two different kinds of information: the
basic conceptual atoms out of which more complex ideas are built, and
structural information that tells you how those atoms are combined and
produces compositional semantics.

An isolating bicategorial language divides those two functions into
two parts of speech. And it probably says something about human
cognition that there exist natlangs with exactly that split, but none
known that are truly monocategorial. Monocategorial languages, on the
other hand, must be inflecting, and divide those functions between
roots (concepts) and inflections (structure). If the morphology is
sufficiently transparent, then such a language can be re-analyzed as
bicategorial simply by changing where you choose to put the word
boundaries.

-l.