A new theory on sentence structure ?

[Mok-Kong Shen]

Could some experts examine the claim in the following URL

http://www.news.cornell.edu/stories/Sept12/SimpleLanguage.html

with the title "Language use is simpler than previously thought"
and tell whether that's true?

M. K. Shen

[DKleinecke]

From where I sit this is one of the dumber non-linguist incursions
into language.

Nathan will doubtless be displeased to see that either the researchers
or the reporter assumes linguistics runs on Chomskian wheels. I quote
-

"In contrast, the hierarchy concept suggests humans have language
thanks only to highly specialized "hardware" in the brain, which
neuroscientists have yet to find."

This, of course, would only be said by someone who was only aware of
Chomsky's Theories. Hierarchy and brain hardware are two different
ideas.

If I understand the history correctly Chomsky's original idea before
SS was string processing (which is what I think the article is
describing) and he discovered it didn't work and changed to structured
models.

But then maybe they have something that eluded me.

[Nathan Sanders]

In article
<9f667b00-c975-4ec0...@ph9g2000pbb.googlegroups.com>,

DKleinecke <dklei...@gmail.com> wrote:

> On Sep 26, 2:17 pm, Mok-Kong Shen <mok-kong.s...@t-online.de> wrote:
> > Could some experts examine the claim in the following URL
> >
> > http://www.news.cornell.edu/stories/Sept12/SimpleLanguage.html
> >
> > with the title "Language use is simpler than previously thought"
> > and tell whether that's true?
> >
>
> From where I sit this is one of the dumber non-linguist incursions
> into language.
>
> Nathan will doubtless be displeased to see that either the researchers
> or the reporter assumes linguistics runs on Chomskian wheels. I quote

Not so much displeased as unsurprised.

> "In contrast, the hierarchy concept suggests humans have language
> thanks only to highly specialized "hardware" in the brain, which
> neuroscientists have yet to find."
>
> This, of course, would only be said by someone who was only aware of
> Chomsky's Theories. Hierarchy and brain hardware are two different
> ideas.

Indeed. It's possible to believe in hierarchical structure in
language without being Chomskyan, but linguistic innateness is a core
Chomskyan principle.

> If I understand the history correctly Chomsky's original idea before
> SS was string processing (which is what I think the article is
> describing) and he discovered it didn't work and changed to structured
> models.
>
> But then maybe they have something that eluded me.

If they're using just plain linear strings, it likely won't work.

If they're incorporating statistical/probabilistic factors, it might
eventually work, but probably not.

The full text of the paper is here:

<http://rspb.royalsocietypublishing.org/content/early/2012/09/05/rspb.2
012.1741.full.pdf+html>

Interestingly, they note:

Unsurprisingly, they seem to be arguing for a slightly weaker claim
than what the news reports imply:

"This is not to say that hierarchical operations are nonexistent, and
we do not want to exclude their possible role in language
comprehension or production. [...W]e expect that it will be possible
to explain much of human language behaviour using just sequential
structure. Thus, linguists and psychologists should take care to only
invoke hierarchical structure in cases where simpler explanations,
based on sequential structure, do not sufŽce."

I've long thought linear order needed more prominence, and even argued
for it in my syntax qualifying paper in grad school.

Nathan

--
Department of Linguistics
Swarthmore College
http://sanders.phonologist.org/

[Arnaud F.]

Le jeudi 27 septembre 2012 04:01:13 UTC+2, Nathan Sanders a écrit :
> In article
>
> <9f667b00-c975-4ec0...@ph9g2000pbb.googlegroups.com>,
>
> DKleinecke <dklei...@gmail.com> wrote:
>
>
>
> > On Sep 26, 2:17 pm, Mok-Kong Shen <mok-kong.s...@t-online.de> wrote:
>
> > > Could some experts examine the claim in the following URL
>
> > >
>
> > > http://www.news.cornell.edu/stories/Sept12/SimpleLanguage.html
>
> > >
>
> > > with the title "Language use is simpler than previously thought"
>
> > > and tell whether that's true?
>
> > >
>
> >
>
> > From where I sit this is one of the dumber non-linguist incursions
>
> > into language.
>

***

The paper in RSP-B sounds like a meta-philosophical short essay written by a teenager parading as scientific discourse.

A.
***

> >
>
> > Nathan will doubtless be displeased to see that either the researchers
>
> > or the reporter assumes linguistics runs on Chomskian wheels. I quote
>
>
>
> Not so much displeased as unsurprised.
>
>
>
> > "In contrast, the hierarchy concept suggests humans have language
>
> > thanks only to highly specialized "hardware" in the brain, which
>
> > neuroscientists have yet to find."
>
> >
>
> > This, of course, would only be said by someone who was only aware of
>
> > Chomsky's Theories. Hierarchy and brain hardware are two different
>
> > ideas.
>
>
>
> Indeed. It's possible to believe in hierarchical structure in
>
> language without being Chomskyan, but linguistic innateness is a core
>
> Chomskyan principle.
>
>
>
> > If I understand the history correctly Chomsky's original idea before
>
> > SS was string processing (which is what I think the article is
>
> > describing) and he discovered it didn't work and changed to structured
>
> > models.
>
> >
>
> > But then maybe they have something that eluded me.
>
>
>
> If they're using just plain linear strings, it likely won't work.
>
>
>
> If they're incorporating statistical/probabilistic factors, it might
>
> eventually work, but probably not.
>
>
>
> The full text of the paper is here:
>
>
>
> <http://rspb.royalsocietypublishing.org/content/early/2012/09/05/rspb.2
>
> 012.1741.full.pdf+html>
>
>

***

The RSP-B are now specializing in publishing crap about language(s) and linguistics with no linguist involved.

A.
***

>
> Interestingly, they note:
>
>
>
> Unsurprisingly, they seem to be arguing for a slightly weaker claim
>
> than what the news reports imply:
>
>
>
> "This is not to say that hierarchical operations are nonexistent, and
>
> we do not want to exclude their possible role in language
>
> comprehension or production. [...W]e expect that it will be possible
>
> to explain much of human language behaviour using just sequential
>
> structure. Thus, linguists and psychologists should take care to only
>
> invoke hierarchical structure in cases where simpler explanations,
>

> based on sequential structure, do not sufÞce."

[Franz Gnaedinger]

Reminds me of the alternative grammar developed by
Pater Rupert Ruhstaller OSB who gave me the first introduction
into modern linguistics in the 1960s. His grammar is based
on functors and arguments. The functors can be visualized
as centers of circles, and the arguments as points on the
circumference of a circle which can become the centers of
further circles - budding circles. Have a look at the drawings
on my web page http://www.seshat.ch/grammar.htm
My opinion is that we need a variety of grammars that
explain different aspects, and a basic definition of language,
here mine from 1974/75:

Language is the means of getting help, support and
understanding from those we depend upon in one way
or another --- and every means of getting help, support
and understanding may be called language, on whatever
level of life it occurs ...

If you start from this definition, language and grammar
gain more dimensions. For example a basic function of
language is to raise and guide attention. This has a grammar
of its own. Body language is included, and again has a grammar
of its own. Etcetera. Each grammar is worth being studied in
its own right, and then plays a role in the concert of grammars,
if I may say so. Consider that paper as one more contribution
to a wider understanding of language, and give up the chimera
of one single grammar that explains all and everything.

[Franz Gnaedinger]

Sorry for the invalid link. Here the correct one:

http://www.seshat.ch/home/grammar.htm

[Arnaud F.]

Le jeudi 27 septembre 2012 08:32:11 UTC+2, Franz Gnaedinger a écrit :

>
> Sorry for the invalid link. Here the correct one:
>
>
>
> http://www.seshat.ch/home/grammar.htm

***

No, it is still invalid.

A.

[Arnaud F.]

Le jeudi 27 septembre 2012 04:01:13 UTC+2, Nathan Sanders a écrit :


>

> <http://rspb.royalsocietypublishing.org/content/early/2012/09/05/rspb.2
>
> 012.1741.full.pdf+html>
>
>
***

There's apparently another interesting paper here on a different topic:

http://rspb.royalsocietypublishing.org/content/early/2012/09/22/rspb.2012.1842.abstract?sid=3a0789f1-f93e-4f7a-a7a7-f073f0351541

I cannot access the pdf.

A.

[DKleinecke]

On Sep 26, 7:01 pm, Nathan Sanders <sand...@alum.mit.edu> wrote:

> Indeed.  It's possible to believe in hierarchical structure in
> language without being Chomskyan, but linguistic innateness is a core
> Chomskyan principle.

My point exactly. Hierarchical structure goes back to the dawn of
linguistics and is imperative if we are going to recognize recursion
(another ancient idea in linguistics). It is almost impossible to do
linguistics without recursion (see Daniel Everett).

But innate language is, so far as i know, believed in only by
Chomskians, and not all of them. However the attempt by Chomsky to
eclipse all other work in linguistics has succeed among the general
public and most people believe that innateness is an article of faith
among linguistics. I do not approve of that.

Concerning sequential analysis it seems obvious that language
processing must be sequential since that's the way it arrives at our
minds. A lot of things can be done without breaking out of that one
sequence but far from all of language. At the very least we need a
pushdown stack to handle back references. I don't know if this has
been investigated but I would guess that humans lose data if the an
internal stack is more than about a half dozen deep. Nor do I know if
we need more than that stack.

The mind seems to have some capability for parallelism and that, I
think, would be the next place to look after the stack.

The hierarchy is too static to be the whole story and perhaps even
handicaps us while we think in terms fo the actual sequence of
language processing.

[Peter T. Daniels]

On Sep 27, 8:15 pm, DKleinecke <dkleine...@gmail.com> wrote:
> On Sep 26, 7:01 pm, Nathan Sanders <sand...@alum.mit.edu> wrote:
>
> > Indeed.  It's possible to believe in hierarchical structure in
> > language without being Chomskyan, but linguistic innateness is a core
> > Chomskyan principle.
>
> My point exactly. Hierarchical structure goes back to the dawn of
> linguistics and is imperative if we are going to recognize recursion
> (another ancient idea in linguistics). It is almost impossible to do
> linguistics without recursion (see Daniel Everett).
>
> But innate language is, so far as i know, believed in only by
> Chomskians, and not all of them. However the attempt by Chomsky to
> eclipse all other work in linguistics has succeed among the general
> public and most people believe that innateness is an article of faith
> among linguistics. I do not approve of that.
>
> Concerning sequential analysis it seems obvious that language
> processing must be sequential since that's the way it arrives at our
> minds. A lot of things can be done without breaking out of that one
> sequence but far from all of language. At the very least we need a
> pushdown stack to handle back references.  I don't know if this has
> been investigated but I would guess that humans lose data if the an
> internal stack is more than about a half dozen deep. Nor do I know if
> we need more than that stack.

As George A. Miller (who died a few weeks ago in his late 90s) put it
in a famous paper, "five plus or minus two."

[benl...@ihug.co.nz]

"The magical number seven, plus or minus two: Some limits on our
capacity for processing information". Psychological Review 63 (2): 81–
97 (1956)

[Peter T. Daniels]

On Sep 27, 11:26 pm, "benli...@ihug.co.nz" <benli...@ihug.co.nz>
wrote:

> 97 (1956)-

! Does 9 turn up in the relevant experiments? Telephone numbers are
the maximal 7, with the extra 3 digits rigidly separated out as the
"area code" even though they are now integral parts of every number
now.

[Arnaud F.]

Le vendredi 28 septembre 2012 02:15:23 UTC+2, DKleinecke a écrit :
> On Sep 26, 7:01 pm, Nathan Sanders <sand...@alum.mit.edu> wrote:
>
>
>
> > Indeed.  It's possible to believe in hierarchical structure in
>
> > language without being Chomskyan, but linguistic innateness is a core
>
> > Chomskyan principle.
>
>
>
> My point exactly. Hierarchical structure goes back to the dawn of
>
> linguistics and is imperative if we are going to recognize recursion
>
> (another ancient idea in linguistics). It is almost impossible to do
>
> linguistics without recursion (see Daniel Everett).
>
>
>
> But innate language is, so far as i know, believed in only by
>
> Chomskians, and not all of them. However the attempt by Chomsky to
>
> eclipse all other work in linguistics has succeed among the general
>
> public and most people believe that innateness is an article of faith
>
> among linguistics. I do not approve of that.

***

The word "innate" is certainly a bad choice,
as it's obvious that language(s) is/are an innate and natural feature of mankind.
There's no instance of a social group of human beings that would not use at least one language constantly, so language is innate: "possessed as an essential characteristic, inherent", according to my dictionary.

But Chimpskyists are certainly experts at twisting the meanings of words, and better at doing that than at explaining how language works.

A.
***

>
>
>
> Concerning sequential analysis it seems obvious that language
>
> processing must be sequential since that's the way it arrives at our
>
> minds. A lot of things can be done without breaking out of that one
>
> sequence but far from all of language. At the very least we need a
>
> pushdown stack to handle back references. I don't know if this has
>
> been investigated but I would guess that humans lose data if the an
>
> internal stack is more than about a half dozen deep. Nor do I know if
>
> we need more than that stack.
>
>
>
> The mind seems to have some capability for parallelism and that, I
>
> think, would be the next place to look after the stack.
>
>
>
> The hierarchy is too static to be the whole story and perhaps even
>
> handicaps us while we think in terms fo the actual sequence of
>
> language processing.

***

What does "hierarchy" describe?

A.

[Franz Gnaedinger]

Rupert Ruhstaller's grammar is the only one I know of
that considers word order. He based his grammar on functors
and arguments. The natural word order is: main functor, main
argument(s), second functor, second argument(s), and so on.
Virgil's Aeneid begins like this: Arma virumque cano ... This
is the actual word order. The natural word order is: Cano
(main functor) arma que virum (arguments) ... Ruhstaller
drew up a diagram, one side representing the natural word
order, the perpendicular side representing the actual word
order, he connected the points, and got a curve of tension,
the peaks of which being the words

(arma) virum Troiae Italiam Lavinia

http://www.seshat.ch/home/gram02.jpg

The first lines of the epic give a summary, and the tension
diagram reveals a summary of the summary, the epic in
its very shortest form, the man who came from Troy to Italy
where he founded Lavinia.

[DKleinecke]

On Sep 27, 11:57 pm, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:

> The word "innate" is certainly a bad choice,

Their choice - the rest of us just sting along.

> What does "hierarchy" describe?

Actually I have been guessing what they meant. It is not, so far as I
know, a standard technical term. But hierarchy is used often these
days to describe what are also called trees. For example XML.

In this situation every element in the structure can potentially have
a set (usually a list) of sub-structures. Each element is the sub-
element of at most one other element, called its parent, and if you
move from an element to its parent and then to its parent and so on
you never get a loop.

This is the structure you get when expand re-write rules and keep all
the intermediate steps. You can if you wish use parentheses to
describe the structure. Linear processing would be equivalent to
erasing all the parentheses before you start.

[Arnaud F.]

Le samedi 29 septembre 2012 04:18:43 UTC+2, DKleinecke a écrit :
> On Sep 27, 11:57 pm, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:

>
> > What does "hierarchy" describe?
>
>
>
> Actually I have been guessing what they meant. It is not, so far as I
>
> know, a standard technical term. But hierarchy is used often these
>
> days to describe what are also called trees. For example XML.

***

yes, but there are several kinds of trees,

for example, if we take "I go home"

trees or stemmas à la Tesnière look like:
go [top of hierarchy]
" "
" "
I home

Which describes how words like I and home depends on the predicate = to go

This is not the same kind of "trees", as using GN GV and all the rest.

A.
***

[Hans Aberg]

On 2012/09/29 10:10, Arnaud F. wrote:
> Le samedi 29 septembre 2012 04:18:43 UTC+2, DKleinecke a écrit :
>> On Sep 27, 11:57 pm, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:
>
>>
>>> What does "hierarchy" describe?
>>
>>
>>
>> Actually I have been guessing what they meant. It is not, so far as I
>>
>> know, a standard technical term. But hierarchy is used often these
>>
>> days to describe what are also called trees. For example XML.
> ***
>
> yes, but there are several kinds of trees,
>
> for example, if we take "I go home"
>
> trees or stemmas à la Tesnière look like:
> go [top of hierarchy]
> " "
> " "
> I home
>
> Which describes how words like I and home depends on the predicate = to go
>
> This is not the same kind of "trees", as using GN GV and all the rest.

You might look at the ASTs (Abstract Syntax or Semantic Tree) used when
constructing compilers:
https://en.wikipedia.org/wiki/Abstract_syntax_tree

The math name for this kind of trees one gets, also when using nested
parenthesizes, is a finite directed and ordered tree. "Directed" means
that the edges are arrows, which is top down in your diagram above, and
"ordered" means that there is a total ordering among the edges of each
node, which is left-to-right above.

Using the notation similar to that of the paper, I might use say prefix
notation f(x_1, ..., x_k) to write a node f with a sequence of edges
x_1, ..., x_k. Then the diagram above becomes "go"("I", "home").

The next question is whether it is useful for describing natural
languages, where the order might shift. It might be better to use
labels. For example, "go"{A: "I", B: "home"}, so that this is the same
as "go"{B: "home", A: "I"}, that is, the order can be shifted - this is
why I sneaked in "{...}" instead of "(...)", looking more like sets or
data structures in C. This is used in some computer languages too, but
from the formal point of view, one just have another data type usually
called a "associative array".
https://en.wikipedia.org/wiki/Associative_array

Hans

[DKleinecke]

On Sep 29, 1:10 am, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:
> Le samedi 29 septembre 2012 04:18:43 UTC+2, DKleinecke a écrit :
>
> > On Sep 27, 11:57 pm, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:
>
> > > What does "hierarchy" describe?
>
> > Actually I have been guessing what they meant. It is not, so far as I
>
> > know, a standard technical term.  But hierarchy is used often these
>
> > days to describe what are also called trees. For example XML.
>

> yes, but there are several kinds of trees,
>

Any of the different trees could be intended. The word hierarchy does
not imply what kind. In fact, it usually doesn't imply a tree - but
the context seems to require such a reading.

[Hans Aberg]

A more general structure is a formal grammar
https://en.wikipedia.org/wiki/Formal_grammar
When implementing a computer language, one attaches semantic actions to
the production rules.

Hans

[DKleinecke]

The question really is what did the authors of a naive article about
language mean by hierarchy. I think any construction more complicated
than a string would fall under their concept of hierarchy.

There are other ways to write compilers (assuming that's what you
meant by implementing).

[Hans Aberg]

On 2012/10/01 01:51, DKleinecke wrote:
> On Sep 30, 1:05 am, Hans Aberg <haberg-n...@telia.com> wrote:
>> On 2012/09/30 06:34, DKleinecke wrote:
>>
>>> On Sep 29, 1:10 am, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:
>>>> Le samedi 29 septembre 2012 04:18:43 UTC+2, DKleinecke a écrit :
>>
>>>>> On Sep 27, 11:57 pm, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:
>>
>>>>>> What does "hierarchy" describe?
>>
>>>>> Actually I have been guessing what they meant. It is not, so far as I
>>
>>>>> know, a standard technical term. But hierarchy is used often these
>>
>>>>> days to describe what are also called trees. For example XML.
>>
>>>> yes, but there are several kinds of trees,
>>
>>> Any of the different trees could be intended. The word hierarchy does
>>> not imply what kind. In fact, it usually doesn't imply a tree - but
>>> the context seems to require such a reading.
>>
>> A more general structure is a formal grammar
>> https://en.wikipedia.org/wiki/Formal_grammar
>> When implementing a computer language, one attaches semantic actions to
>> the production rules.

> The question really is what did the authors of a naive article about
> language mean by hierarchy. I think any construction more complicated
> than a string would fall under their concept of hierarchy.

They do not seem to be very specific, just quoting some other articles.

> There are other ways to write compilers (assuming that's what you
> meant by implementing).

I had in mind not relying entirely on a grammar, but using the semantic
actions as well to pin down the language.

Hans

[DKleinecke]

Without semantics a language is pretty useless. But while we all have
come to agree (more or less) about formal syntax there seems to be no
agreement on the best way to represent the semantics.

For example I write compilers to what is technically a subset of C and
then assume the C compiler does all the optimizing dirty work. I feel
compilers are easy to write if it weren't for all the special cases
involved in implementing the language on a real computer. That part
is an exercise in pure boredom.

[Hans Aberg]

> Without semantics a language is pretty useless. But while we all have
> come to agree (more or less) about formal syntax there seems to be no
> agreement on the best way to represent the semantics.

Right, though there is denotational semantics.

> For example I write compilers to what is technically a subset of C and
> then assume the C compiler does all the optimizing dirty work.

For example, LLVM admits optimizations that are not through C.

> I feel
> compilers are easy to write if it weren't for all the special cases
> involved in implementing the language on a real computer. That part
> is an exercise in pure boredom.

That is legacy; comp.compilers had in March a thread about lack of
progress in programming language design.

[DKleinecke]

I am pretty old-fashioned about computer language. Although I would
prefer to code in assembly language I settle for 1990 ANSI standard C
with one improvement - declarations floating in blocks. I am unsure
that any "progress in programming language design" is a Good Thing.

[Hans Aberg]

The 2011 revisions of C and C++ are rather quickly implemented now.

[DKleinecke]

Sure. But I am old-fashioned enough to deplore the innovations. There
are even things in 1990 ANSI I deplore and do not use - for example,
FOR statements. FOR statements are not, if I remember correctly, in
the oldest surviving version of C - but that's not my objection. I
think they add unnecessary complexity to code.

[pauljk]

"DKleinecke" <dklei...@gmail.com> wrote in message
news:8bad5df4-b46c-4dee...@n7g2000pbj.googlegroups.com...

> On Oct 3, 9:48 am, Hans Aberg <haberg-n...@telia.com> wrote:
>> On 2012/10/03 02:23, DKleinecke wrote:
>>
>>
>>
>>
>>
>>
>>
>>
>>
>> > On Oct 2, 3:46 am, Hans Aberg <haberg-n...@telia.com> wrote:
>> >> On 2012/10/02 04:43, DKleinecke wrote:
>>
>> >>> On Oct 1, 1:42 am, Hans Aberg <haberg-n...@telia.com> wrote:
>> >>>> On 2012/10/01 01:51, DKleinecke wrote:
>>
>> >>>>> On Sep 30, 1:05 am, Hans Aberg <haberg-n...@telia.com> wrote:
>> >>>>>> On 2012/09/30 06:34, DKleinecke wrote:
>>
>> >>>>>>> On Sep 29, 1:10 am, "Arnaud F." <fournet.arn...@wanadoo.fr> wrote:

>> >>>>>>>> Le samedi 29 septembre 2012 04:18:43 UTC+2, DKleinecke a �crit :

Your avoiding the FOR statements reminded me of an article
I read many, many, many years ago when I was in a small group
of people writing an Algol 60 compiler. The magazine (it may have
been CACM, but I am not sure anymore) article advocated replacing
all Fortran FOR statements with ON GOTOs and calculated GOTOs.
Jiminy crickets I thought to myself, calculated GOTOs, that guy
must be nuts. Towards the end of the article the author embarked on
discussing the virtues of the proposed COME FROM. Wow, wow,
wait a minute!!! I checked the date of the issue, and yes it was
the 1st of April.

pjk

[Hans Aberg]

Have you though of switching to B? Not even Bison will support these
outdated versions of C.

[DKleinecke]

Not at all sure which versions of C you as disparaging. If you mean
1990 ANSI it is still a subset and newer software - like Gnu CC - work
just fine. There is no rule saying you must use all features of a
language. If you mean the original version, it had other defects that
the 1990 version almost entirely cleared up.

[Hans Aberg]

It is K&R C that will not be supported in Bison 2.6. This means that one
cannot expect a compiler only accepting this language to compile a Bison
generated parser file.

[DKleinecke]

Which matters, of course, only if one uses Bison. I don't.

I haven't checked in detail but I believe 1990 ANSI corrected several
genuine problems with K&R and they are not identical.

[Hans Aberg]

The problem is only if you expect your subset to be supported.