[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >

[Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: [engelang] Re: Self-segmenting words & the treatment of names

Mart�n Bald�n, On 14/05/2006 21:28:

--- In engelang@yahoogroups.com, And Rosta <a.rosta@...> wrote:

Now that I've seen it, my thoughts are as follows:

The syntax as described brackets words together into groups. But if

you examine the structure of a logical representation, it consists of
predicate--argument relations. So what one asks from a loglang syntax
is not that it group words together but that it represent
predicate--argument relations.

Brackets are used in Lisp (and also in knowledge representation
languages like SUO-KIF and CycL) in order to delimit terms. The syntax
  of a sentence in these languages is, loosely speaking:

(predicate arg1 arg2 arg3 ...)

where "predicate", "arg1", "arg2" are called terms and they can, in
turn, be other parenthesised expressions. 

Setting apart the matter of the predicate being itself a phrase, you can do without left brackets, can't you. And if predicates have a determinate number of arguments, then you can do without right brackets too.


Notice that, when we build complex expressions from simpler ones, the
inner markup of the constituent expressions is preserved,
and only the outest parenthesis of each one of them is changed. This
is a form of encapsulation, which is one of my top priority design
criteria.

What does 'encapsulation' mean, in this context?


It means that, when you build bigger expressions from smaller ones,
you don't have to modify the smaller expressions internally. If you
make the hierarchical markup number be greater for inner parenthesis
than for outer ones (which I call the "top-down" approach), then you
are not providing encapsulation, because you have to change the
internal form of an expression so as to use it as a building block for
another expression. This doesn't mean that you loose any information,
it only means that you can't use it litterally. 
On the other hand, in the "bottom up" approach, which I'm using, the
markup of the constituent expressions is unchanged (except their
outest parenthesis, but you can consider that they are not part of the
expression), and it's the bigger expression which has to adjust its
markup accordingly.

The drawback is, arguably, that you have to know the hierarchical
level of every term before you can use them to build a sentence. To
this, I can say that you only have to do that if you want to fuse
contiguous parenthesis, but you can ignore this at the expense of not
fusing them. 

You could dispense with 'numbered' brackets, and the problem would go away, wouldn't it?

Or you could switch to reverse polish.


Besides, it seems to
me that glottal stops are best used between vowels, and difficult to
combine with consonants, 

Many accents of English have glottal stops (as realization of /t/) in consonant clusters.

--And.