Hi Christophe,

This is only a formalism created by Chomsky in order to explain how would a
human brain parse sentences. It is part of Chomsky's idea of a Universal
Grammar.

The context-free languages can be parseable by using a pushdown automaton
(a kind of stack).

As one cannot implement cross-dependencies by using a stack (as it is a
LIFO structure - last in/first out). On the other hand, if a queue were
used as the main structure of the automaton, then cross-dependencies would
be OK (a queue is a FIFO structure - first in/first out), but one would not
be able to parse languages whose dependencies are equidistant from a given
point.

Example:
In German you can have: Johann sagt, dass wir die Kinder (dem) Hans das
Haus anstreichen helfen lassen wollte.
Using the space (" ") between "das Haus" and "anstreichen" as a central
point, one can see that there are dependencies between:
- "anstreichen" and "das Haus" (accusative-regency verb and accusative
noun);
- "helfen" and "(dem) Hans" (dative-regency verb and dative noun);
- "lassen" and "die Kinder" (accusative-regency verb and accusative noun).

Cross-dependencies occur in Zürich German: Jan säit das mer d'chind em Hans
es huus haend wele laa hälfe aastriiche.
Using the space (" ") between "es huus" and "haend" as a central point, one
can see that there are dependencies between:
- "laa" and "d'chind" (accusative-regency verb and accusative noun);
- "hälfe" and "em Hans" (dative-regency verb and dative noun);
- "aastriiche" and "es huus" (accusative-regency verb and accusative noun).


As I visualize, imagine human brain works as either a queue or a stack to
parse sentences is not weird. It would be weird if it would be capable of
swap between using queues or using stacks depending on the sentence
context. (Queue or stack is only a matter of definition. What would destroy
this theory would be a language that can use both - Some languages using
queues, and some using stacks, would seem to destroy the theory of a
"universal grammar" - if there should exist a "universal grammar" it would
have both... or none of them).



Anyway, due to Shieber's discovery (about Züritüütsch use of
cross-dependencies), a new class of formal languages was created, the
mildly-context-sensitive languages. These languages are defined based on
three constraints, and the most important one (in my opinion) is that the
languages that belong to this class must have "constant growth".
If you read section 4 of paper (
http://wwwhomes.uni-bielefeld.de/mkracht/html/suffixaufnahme.pdf) - in page
7 - you'll see taht he uses what he calls a "word marking English". He uses
"gen" to denote the genitive case. The amount of genitive affixes does not
constantly, but at a quadratic rate.

He is trying to write "the son of the son of the son of the son runs".
In English, the amount of "of"-words grows constantly. "The son runs" (0),
"The son of the son runs" (1), "The son of the son of the son runs" (2),
"The son of the son of the son of the son runs" (3) and so on;
In Latin it also grows constanly. "Filius currit" (0), "Filius filii
currit" (1), "Filius filii filli currit" (2), "Filius filii filii filii
currit" (3) and so on;
On the other hand, in Marked Word English it does not grow constanly. "Son
run" (0), "Son songen run" (1), "Son songen songengen run" (3), "Son songen
songengen songengengen run" (6) and so on.


What I proposed is that the surdéclinaison would make the language affixes
grow at a higher rate than a constant one. Maybe, by knowing the difference
between "Suffixaufname" and "Surdéclinaison", you could be able to prove if
"Suffixaufname" is non-context-free/non-mildly-context-sensitive. If
"Suffixaufname" is mildly context-sensitive or not has created a big debate
between Michaelis and Kracht, and Joshi.

Michaelis and Kracht paper (Semilinearity as a syntactic invariant):
ftp://ftp.math.fu-berlin.de/pub/usr/kracht/sprache/semilinear.ps
Joshi paper (Semilinearity is a syntactic invariant: A reply to Michaelis
and Kracht (1997)): <http://people.umass.edu/bhatt/papers/li-semilinear.pdf>
http://people.umass.edu/bhatt/papers/li-semilinear.pdf



On Mon, Apr 16, 2012 at 12:19 PM, Christophe Grandsire-Koevoets <
[log in to unmask]> wrote:

> On 12 April 2012 19:26, Logan Kearsley <[log in to unmask]> wrote:
>
> > On 10 April 2012 13:14, Armin Buch <[log in to unmask]> wrote:
> > > Hello Conlangers,
> > >
> > > this is going to be very linguistic and mostly uninteresting - unless
> you
> > > want to know how to make your conlang even more different from
> natlangs.
> > >
> > > I was just asked an interesting question: Are there any
> context-sensitive
> > > conlangs?
> >
> > Christophe's posts on Suffixaufnahme led me to reading this paper (:
> > http://wwwhomes.uni-bielefeld.de/mkracht/html/suffixaufnahme.pdf)
> > which mentions that Swiss German is context-sensitive and argues that
> > Old Georgian is context-sensitive. Without actually going through the
> > mathematical arguments, since I don't know Moten to be able to do
> > that, I would guess that Moten is probably context-sensitive for
> > similar reasons.
> >
> >
> Hi everyone!
>
> Wow, I go away and offline for a week (I was in France to celebrate my
> grandmother's 90th birthday :) ), and I come back discovering I've been
> mentioned not once, but twice in the same thread! I just couldn't not react
> :) .
>
> However, I couldn't tell you whether Moten is or isn't context-sensitive,
> as I am not sure I understand what it means, or for that matter the whole
> subject of formal grammar. Indeed, I used the paper mentioned above as one
> of the foundations of my LCC4 presentation, but I deliberately avoided the
> formal part of it, if only because my understanding of this area of
> mathematics has always been rather poor (I'm more of a Calculus kind of guy
> :P ). If anyone could explain (in layman terms) what this is all about, and
> what the difference between context-free and context-sensitive languages
> concretely *is* (try not to use parse trees either: I understand them, but
> I'd like a more concrete explanation), then I'd be very grateful. I've
> tried reading the links already given, but the formalism goes a bit over my
> head, and the absence of concrete examples doesn't help.
>
> Also, maybe someone with a better grasp of things wants to have a look at
> the corpus of Moten I've already used in my blog posts and see whether they
> give any clues as to the context-freeness or -sensitiveness of Moten
> grammar, although I don't know whether one can make conclusions already
> based on the examples I gave.
>
> Finally, if this is really only about whether in parse trees branches cross
> each other or not, then I'd like to know what the big deal is. I'm pretty
> sure if you parsed 90% of the texts I had to understand and translate in
> Latin class, you'd see lots of branches crossing. Cicero had some weeeird
> speeches out there, and Classical Latin poetry can be frighteningly
> hermetic! :P Does it mean Latin grammar is context-sensitive? I wouldn't
> know. Would knowing give us some deep understanding of the Latin language
> that we would otherwise lack? That's my question. If so, then I'd be very
> happy to know what that understanding is, concretely speaking :) .
>
> Thanks for enlightening this layman!
> --
> Christophe Grandsire-Koevoets.
>
> http://christophoronomicon.blogspot.com/
> http://www.christophoronomicon.nl/
>