Language without grammar


    Sponsored Links


  • 1. Ostensible Definition
    Ostensible Definition ~v~~ Let's suppose we wish to define X but we don't know how to do it exhaustively. In effect then we say that X is as X does. In other words instead of saying X is "thus and such" we say X = f(X). The difficulty however then becomes the definition of f(X). How do we define f(X)? Are we simply to say that f(X) = f(f(X))? Obviously there is no way to define f(X) without defining X first and definitions for X which rely on the behavior of X are unavoidably circular since implicitly they rely on behavior for something which is not defined. ~v~~

Language without grammar

Postby JXStern » Thu, 16 Dec 2004 05:40:07 GMT

was: Useless thought of the day: Language Compression

On Tue, 14 Dec 2004 04:12:49 +0000 (UTC), Neil W Rickert


I think we agree here, but there's a lot of quibbling still to do
about the terms of art.  What's a language, a protocol, a program, an
interpretation, a fact, a proposition, an intenxion ...

I think that defining language as that which a computational agent
uses to effect a result, is pretty close to what people view
computational languages as about anyway.  If you start with a language
of thought, focus on agency, and stick to nominalism, I think you're
pretty much there.  The whole thing (think even standard computer
programming) is very pragmatic/extensional, once you stop trying to
cram it into a platonistic/logicistic framework.

Joshua Stern

Re: Language without grammar

Postby Neil W Rickert » Thu, 16 Dec 2004 11:47:15 GMT

Hash: SHA1

JXStern < XXXX@XXXXX.COM > writes:

You could spend a long time arguing about those.

I don't want to start with a language of thought.  I doubt that one

If we use "thought" to describe what happens in a computer, then
there is no language of thought there.  The computer deals with
representations.  These representations are complex data structures
in memory, and not anything that looks like a linguistic

We can have a representation system of thought, but that
representation system need not be a language.  In fact it probably
should not be a language.

Version: GnuPG v1.3.91 (SunOS)


Re: Language without grammar

Postby Wolf Kirchmeir » Thu, 16 Dec 2004 12:17:59 GMT

Grammars are devised by linguists. We should not be talking about the 
mess that passes for grammar in the typical English-language school room.

Grammars as devised by linguists are systematic descriptions of how a 
language expresses what all languages express, so that comparison with 
(all!) other languages is implicit. Talk about grammars of natural 
languages by people without linguistic training is 99.9% beside the point.

There is occasional useful talk from computer programmers, who have to 
know the "rules of syntax" of their programming codes. But don't for a 
moment think that such rules have anything other than  metaphorical 
relation to actual grammars of actual languages.

BTW, to the extent that the phrase "communication exchanges being made 
under an ad hoc fault tolerant protocol" makes sense, it refers to 
language behaviours. Fine phrasing, but empty if it's not based on 
observation and experiment - neither of which are particularly easy when 
it comes to language behaviour. Just because you can express yourself 
effectively in some language doesn't mean you know how you do it, nor 
how anyone else does it, either. The only practical study of how 
language works that most people have been exposed to is the study of 
literature, a comment that is not intended as a compliment.

Re: Language without grammar

Postby JXStern » Thu, 16 Dec 2004 12:55:10 GMT

On Wed, 15 Dec 2004 02:47:15 +0000 (UTC), Neil W Rickert

My Pentium has a "language of thought" in its binary codes.  That's
what I mean, and that's all I mean, by LOT.  Frege meant something
both more and less with his logicism.

There is nothing *but* an LOT there, even if the computer doesn't
think as such.

I am even more dubious about the utility of the category
"representation" than you are about "grammar".

I think the representation is as epiphenomenal as you seem to feel the
grammar is.  I think the grammar is harmless and mechanical, not
sufficient and certainly not essential to (or from) meaning, but there
all the same, contingently, nominalistically, and last but not least,
physically.  Whatever role representation may have rides on top of
that, and can be as solid on top of that as anything, but I just don't
know how solid that can be.


Re: Language without grammar

Postby Rob Freeman » Thu, 16 Dec 2004 13:42:39 GMT

I don't mind the "without grammar" moniker as a starting point. But
only as a starting point. I would quite like to keep grammar as a label
for the ad-hoc order I think it is the task of language to find.

So I see us starting from a position of no grammar, like you, but I
would like to think of the patterns we find along the way (one or
another) as grammar.

You also made the point that is is difficult to see how communication
proceeds in a system of this kind, without any underlying logical
representation. This is a leap I agree, because we are turning things
around. Language is no longer the expression of logic, instead it is a
constant search for new logic.

I hinted at the way I see it working in the message you replied to. I
see communication now as being an "organization into the same grammar".

For this to work two individuals would have to start with roughly the
same experiences organized in roughly the same way, and allow some
(new?) language event to organize the examples in their heads into the
same (new) "grammar". (Note: the change in organization associated with
a "new" grammar might be very small, say just that associated with a
change of word order, like "Jack hit Jill" instead of "Jill hit Jack".)

Mostly we have the same experiences, so mostly we would have the same
grammar, and mostly it would work.

Conceptually it is a bit confusing because such a "representation" for
the meaning of a sentence (an organization of examples) is less compact
than the sentence and we are not used to that. It seems like we have no
representation for the meaning. We can't put a label on a sentence and
say "This is what it means." It is as I've said in the past a lot like
a Web search engine. A search key organizes the documents on the Web
into a certain order. The order produced by the search can be thought
of as the "meaning" of that search, and is quite useful, but there is
no more compact label for that order than the search key itself (and
the data-base it operates on.)


Re: Language without grammar

Postby AngleWyrm » Thu, 16 Dec 2004 13:45:35 GMT

What if the constructs we know as language are simply an memory device to
reconstruct an electrochemical firing sequence in the listener? What if the
process of communication is no more than producing an echo in a similarly
constructed biochemical receiver?

An interesting conversation might be one that first recognizes an obvious path,
and then travels down refreshingly unused "perspectives" from the listener's
point of view.

Re: Language without grammar

Postby David Longley » Thu, 16 Dec 2004 20:31:54 GMT

In article < XXXX@XXXXX.COM >, Rob 
Freeman < XXXX@XXXXX.COM > writes
The first below is worth sitting through patiently if one hasn't done so 

< http://www.**--****.com/ ;

Some brief additional background:

< http://www.**--****.com/ ;
< http://www.**--****.com/ 

For the other chair (the last contribution, another Chomsky basher):

< http://www.**--****.com/ ;
David Longley

Re: Language without grammar

Postby lesterDELzick » Fri, 17 Dec 2004 00:53:31 GMT

On Tue, 14 Dec 2004 22:17:59 -0500, Wolf Kirchmeir

So, what's the point.

So much for comparative philology.

Regards - Lester

Re: Language without grammar

Postby lesterDELzick » Fri, 17 Dec 2004 00:57:56 GMT

On Tue, 14 Dec 2004 20:40:07 GMT, JXStern < XXXX@XXXXX.COM >

Language without grammar is knowledge without structure. If things are
known in relation to one another, there is structure in the knowledge
and grammar in the structure.

Regards - Lester

Re: Language without grammar

Postby patty » Fri, 17 Dec 2004 01:59:25 GMT

What does "linguistic representation" mean to you?

What is a "representation system" such that is it not linguistic ?


Re: Language without grammar

Postby Neil W Rickert » Fri, 17 Dec 2004 07:12:36 GMT

Hash: SHA1

JXStern < XXXX@XXXXX.COM > writes:

Well, okay.  I think you are a little too generous in what you will
call a language of thought.

To me, a langauge is something to be used in a serial communication
channel between two agents.  The stream of bits in a computer doesn't
fit very well.  It is really a bunch of communications between a
number of agents, which happen to be multiplexed over a common bus.

I will admit that the term lends itself to a lot of misuse.

Version: GnuPG v1.3.91 (SunOS)


Re: Language without grammar

Postby Neil W Rickert » Fri, 17 Dec 2004 07:14:49 GMT

Hash: SHA1

patty < XXXX@XXXXX.COM > writes:

Among other things, "linguistic" implies that it involves an ordered
sequence of symbols.

Images would be an obvious example.

Version: GnuPG v1.3.91 (SunOS)


Re: Language without grammar

Postby Neil W Rickert » Fri, 17 Dec 2004 07:32:45 GMT

Hash: SHA1

"Rob Freeman" < XXXX@XXXXX.COM > writes:

I don't have a problem with that. Certainly we use grammar in the
ways that we give organized accounts of language. I was mostly
arguing against grammar as part of the causal structure of language.

I will 'fess up. I am not a true believer in the religion of logic
worship. That is to say, I see logic as a useful tool -- no more, no
less. I do not see it as part of the fundamental structure of

Here is an example of what I had in mind.

Imagine a video camera, and a television display. There is a signal
transmitted between the camera and the display.

I some sense, the video camera has a pixel representation of the
scene being recorded. And that representation is meaningful, in the
sense that it is causally connected to the real world scene.

Likewise, the television display has, in some sense, a pixel
representation. And that is meaningful in the sense that it is
causally connected to the displayed image.

The signal between the camera and the display is a serial
transmission of symbols. I say that it is inherently meaningless.
It follows a protocol to transfer representations from camera to
receiver. But it can be used regardless of whether the camera
representation is meaningful. It isn't the job of that serial
transmission to tranfer meaningful content. Its job, rather, is to
update the receiver representation to match the camera
representation. The meaningfulness in at the end points, not the
intermediate serial data transmission.

If we can pretend that the receiver was conscious, then it would have
the experience of receiving these serial transmissions and as a
result having meaningful content available. Thus it might seem to
the conscious receiver that the serial channel carried meaning. But
that would only be its subjective impression. The proper objective
view should see the meaningfulness as being provided at the two end
points (camera and receiver), but not as anything transmitted.

Version: GnuPG v1.3.91 (SunOS)


Re: Language without grammar

Postby JXStern » Fri, 17 Dec 2004 11:10:06 GMT

On Wed, 15 Dec 2004 22:32:45 +0000 (UTC), Neil W Rickert

Amen and amen.


Re: Language without grammar

Postby JXStern » Fri, 17 Dec 2004 11:13:15 GMT

On Wed, 15 Dec 2004 22:12:36 +0000 (UTC), Neil W Rickert

It's not that I was looking to be that generous, it just seems to me
that's how things work out.

Yes, but it just so happens that particular sets of bits that I use,
prove useful to others who also are running Pentium-architectures.
It's not a communications as such, but it is a commonality that we can
depend on in understanding how agents that are physically built as
instances of a kind (human, Pentium), generate similar behaviors from
similar internal states.  It's too easy to take this for granted, it's
a point that needs serious and frequent consideration, and if given
this, it carries a lot of weight.

Joshua Stern

Similar Threads:

1.Re Language without grammar

This message did not show up on my server. I've found it
on Google groups today. But for some reason Google does not allow
me to reply to any message. It says "unable to retrieve this
message." So I tried to post my reply separately. And then
Google refuses the subject lines starting with "Re:..."  

"JXStern" < XXXX@XXXXX.COM > wrote
>On Tue, 28 Dec 2004 19:25:53 -0500, "Pierre-Normand Houle"
>< XXXX@XXXXX.COM > wrote:
>>> When I say there are levels, these are levels of analysis only,
>>> is still just a single computer with a billion or so little
buckets of
>>> electrical charge and whatnot sitting on my desk.
>>First, it seems to me that each distinct level of analysis, in as
much as it is
>>perspicuous, should carry some ontological commitments. (Quine : "To
>>is to be the value of a bound variable.") So when you say "there is
>>do you mean that there aren't a computer case, stickers, circuit
boards, wires,
>>and so on? Are there only electrons, quarks and some such things?
But then
>>in what sense can you say that these things are sitting on your
>In a simple, intuitive, and ultimately completely rigorous,
>*particular* way.  I'm not suffering any problems with rabbithood
>here.  When I go to the store to buy a new computer, the clerk
>collapse in a puddle of existential angst about it.  I'd say it
>behooves philosophy to be able to deal with the real world with equal

You were suggesting that there were both narrow and wide contents
and that these belonged to different levels. I proposed that
contents are always, in different senses, wide. Your answer,
which I do not quite understand, is that those levels you were
talking about are merely levels of analysis for there is only
one single thing we are talking about: a computer which is merely
a bunch of physical parts.

But why not say that different levels of analysis reveal different
sorts of *functional* units witch have different sorts of contents?
After all, it is symbols or functional states that have contents,
not mere bunches of electrons or atoms. So the computer is one
functional unit. The robot that is being controlled by the computer
could be another functional unit with intentional (or functional)
states that are not the same as the functional states of the
computer that is controlling its behaviour. The discrete physical
states that the computer itself transitions between--as characterized
at the lowest intelligibly functional levelight correspond roughly
to CPU + memory states. But these very same token computational
states (and not their types!) correspond, at another level, to token
states (or parts of token states) of the operating system. At yet
another level these correspond to states of some virtual machine,
and then to states of some robot-controlling program that runs on
that virtual machine. Now, it seems to me, you could have some
perfect understanding of the functional architecture of some one
of these levels while being completely ignorant of all the levels
below (implementation) and above (emergent levels). Yet you would
know all you need know to ascribe relevant contents to those states.
These would in all cases, (except maybe for the lowest, physical,
level) be wide content for they would require knowledge of the
functional organisation at this level and of the conventional
meaning assigned to the various symbol types of the language being
operated on at that level. 

But there is a sense, also, in which all these contents, at various
levels, are narrow. This is the sense in which the states are
"blind" to the meaning associated to the type symbols at all
the higher levels which their own level "implement" or "realize"
in some wider context.

Another reason I do not understand your reply is that you claim
to be a nominalist. But you also claim that the different "levels"
are mere "levels of analysis" except for *one*: the physical level.
Why is not the physical level also a mere a level of analysis?
Such a position would be physicalism, which admits of "real"
physical universals only, not nominalism. I put a similar objection
to Rickert's view of computational processes as mere attribution
to real physical systems, for his commitment to reductive
physicalism is not explicit either.     

>I'm sure that's just what I am trying to avoid.  Yet, the levels do
>adjoin each other and have to be discussed in the same paragraphs in
>order to sort their issues.
>Everything I'm saying here leads me to assert that my statement here,
>that the program chooses its state, is the only perspicuous manner of
>handling the issues involved.  At a slight cost in anthropomorphism,
>the program has agency in my view, but without commitment to qualia
>free will or such (yet) higher-level properties, or phenomena, or
>capabilities, or whatever category you'd place qualia in anyway.

You've already made a commitment to "choice" so why avoid the other
intentional and phenomenological notions? The program is deterministic
so how can you say that it chooses? 

>> It conflates the intentional level with the
>>software level. We can, following Dennett, consider the possibility
of a chess
>>computer "choosing to mount a king side attack." But the program
>>the intentional level at which such attributions are made makes no
>>It only proceeds according to deterministic software instruction.
>How else would one ever make a judgement?

Though being sensitive to norms of rationality, to which deterministic
algorithms are not, and also through valuing some ends. It seems to
us that the system values victory (but this was really the goal of the
programmers) and maybe the system is genuinely sensitive to higher
order features of chess board positions that warrant the adoption
of definite strategies. But these have not been forecast by the
and neither is the low level program "concerned" with them. They
at the system level. 

>>Compare the system reply to Searle's Chinese room thought
>The system "makes the decision".  I find structural problems with the
>Chinese Room plus or minus the systems reply, but it remains a
>wonderful intuition pump, and I'm comfortable asserting this position
>regarding it.

OK, but the system which makes the "decision" isn't the program. It is
the whole emergent system that is sensitive to the fact that now is
time to launch a king side attack. The program is only subvening this
intentional process that *we* discern in the operation of the system.
The program itself does not picture or represent the chess board but
only a tree of meaningless "moves" that it selects according to some
deterministic tree pruning algorithm. What emerges from that, one
level up, is of no concert to the program. It is the behaviour of
the system that supports primitive intentional attributions. (The
attributions are primitive because the system is still quite brutish.
If you deviate only slightly from the structure of a formal chess
game I/O relationships, the program is lost: it become idle or
garbage outputs from then on.)

For all you know, there could be such a program running in you head
when you play chess. As an upshot, this or that move strikes you as
promising. But this is an emergent feature. You may have no way of
knowing why some move look promising and some other move looks bleak.
Some advanced neurosurgery could replace a bunch of tissue in your
head that is responsible for chess pattern recognition and put
a module implementing three pruning algorithm in its stead (assuming
this were feasible). Some other better of worse chess moves (or the
same moves) would strike you as good, but you yourself, the whole
embodied brain, would still be the agent of those move, assuming
you are committed to the game and rely on some such
causally grounding the correctness of your intuitive judgments.
But those "modules" in your head could still be deterministic and
blind. Where they operate is at least one level down relative to the
intentional level at which you appear (even to yourself) to emerge
as an agent.       

>>The sense in which the programmers decided in advance the rooms
every move
>>is the sense in which they provided the deterministic program.
>If I build a bomb, I am NOT deciding in advance every move of the
>fragments the bomb will produce.

But the bomb does not decide either. Nobody decides. It all goes
according to non-intentional physical laws. 

>Programmers may tell themselves they
>are building something better than bombs, and I hope they are, but
>metaphysically I'm afraid they fall into similar categories.

Indeed. They might not know in advance what output the program will
produce. The system (such as the robot it controls, or the person
whose brain it emulates, say) might be too complex for its being
possible to predict what will be seen as emerging from its operation
in the world. But the program does not decides either what will be
seen as emerging in the world as a result of its individual "actions"
in the flow individual I/Os. It no more *knows*, or need to know,
what happens in the world than a CPU needs to know what operating
system is being ran un it. CPUs are blind. So are algorithms, at
all levels.

>From a
>God's eye view, everything may be deterministic,

But also from a human physicist's point of view.

>but from a pragmatic
>view we tend to lack context and computing power, and perhaps quantum
>uncertainty and deterministic chaos undermine our folk theories of
>determinism anyhow, and in its own way computation shows yet another
>set of limits to determinism. It is absolutely crucial to a proper
>understanding of computation to realize that whatever the programmer
>puts on paper (as if programmers put programs on paper anymore!) has
>no causal connections until the computer runs that program in
>context it finds itself in.  Even if the programmer did successfully
>analyze all possible contexts and execution paths and all that the
>program does was indeed well anticipated, the actual performance of
>that computation by the agency of the computer is a new thing in
>itself and constitutes whatever it is we are constituting - a
>"decision" to move a chess piece, etc.  This is an instrumentalist
>view of things, and it is exactly what I am asserting.

I agree that it is a new agency that we could be dealing with at the
highest level of behaving in the world. But at this level, it is no
longer a program instance or a computer we are dealing with, but
some new entity whose meaningful "I/Os" supervene not only on
the computer which is a part of it and also on social/environmental
features that determine a shared context in which only intentional
attributions make sense.   

>> The sense in which
>>the room "chooses" chess moves is the sense in which there is an
emergent level
>>of sensitivity to rational norms (in this context this involves a
desire to win chess
>>games and a grasp of the necessary means to that end) but this is a
distinct level.
>>In Dennett's terminology, it's the difference between the design and
>>levels. Although Dennett himself would tend to view these as
different stances
>>on the same system this merely obscures the fact that they are
genuinely distinct
>I'm not sure I ever see an "intentional" level in this chess example.
>I'm not sure I see "intention" as anything more than a folk theory to
>be assembled out of more fundamental items.  This sort of overlaps
>Dennett's talk of "stance", but does not overlap it at every point.

This stance or interpretation it is not intended as a folk "theory"
for its main function is not to predict or explain meaningless
behaviour (mere bodily motions or meaningless I/Os) but rather
to disclose behaviour, or language, *as* meaningful expressions
of intentional states and attitudes. Of course, you can eliminate
the concepts of this "folk theory" (common sense psychology) and
still predict "behaviour" from a physical or neurophysiological
stance but then you would eliminate the explanans with the explanandum
(assuming you are interested in such things as proper attributes
of human beings, or intelligent robots, and not just spatial
movements of macroscopic lumps of flesh or metal.)    

>Again, thank you for some references I will have to chase down.
>Meanwhile, I have taken a look at Putnam's "Why Functionalism Didn't
>Work", 1988 in R&R, which refers back to his, "Philosophy and Our
>Mental Life", 1973 collected in "Mind, Language and Reality".  1973!
>And with a little poetic license, I think Neil had it about right.
>Putnam rejects functionalism because he cannot find a lawlike
>relationship between physical state and cognitive state to match that
>between temperature and mean kinetic energy.  Well, I think he is
>wrong, on computation, on the metaphor, and even on the philosophy of

Why? He gave three desiderata for some successful reduction. Is there
one you disagree with?

>I think he is mixing levels.

How so?

>And it's not that I think that
>functionalism does work - I agree it does not!  But quite obviously,
>something *like* functionalism is going on in that we have multiple

It seems to me that it is also Putnam's opinion that something like
functionalism is true. His internal realism is something like
functionalism in some respects. It abstracts to everything that is
not essential to some core Wittgensteinian form of life, including
some bundles of language games, and the essences these must supervene
on in order to fix the references of our thought about particulars
and natural kinds. This view leaves room for enough "multiple
realizability" that we can come to understand each other, communicate,
rationally argue, empathise, contradict each other, and so on, despite
each of us having radically different historical backgrounds, sets of
goals and even metaphysical beliefs about the world. (Not that I quite
endorse internal realism, myself, but something like it ;-) 

>as we obviously do in any space of Turing machines.
>But then, we also have multiple realizability in the case of
>temperature!  Different beakers of water, or a beaker of water and a
>beaker of alcohol, or whatever, and this does NOT defeat the
>generality of the idea of temperature!

This is also Putnam's opinion. Mean kinetic energy is multiply

>  1973!  Taking a historical
>turn here, we find Putnam revising his own earlier proposals at the
>same period when analyticity was generally falling from favor.  'nuff
>said about that, I think.

I apologise, I don't get your drift. What's the link with analyticity?

2.Language without grammar

was: Useless thought of the day: Language Compression

On Tue, 14 Dec 2004 04:12:49 +0000 (UTC), Neil W Rickert
<rickert+ XXXX@XXXXX.COM > wrote:
>I am arguing that nature had no need to provide us with a language.
>All it needed was a protocol.  And what we see as natural language
>are really the communication exchanges being made under an ad hoc
>fault tolerant protocol.  The putative grammar is something invented
>by the grammarians and philosophers -- a useful tool for studying
>some of the regularities of natural language expression, but not
>anything that is a real part of the natural world.

I think we agree here, but there's a lot of quibbling still to do
about the terms of art.  What's a language, a protocol, a program, an
interpretation, a fact, a proposition, an intenxion ...

I think that defining language as that which a computational agent
uses to effect a result, is pretty close to what people view
computational languages as about anyway.  If you start with a language
of thought, focus on agency, and stick to nominalism, I think you're
pretty much there.  The whole thing (think even standard computer
programming) is very pragmatic/extensional, once you stop trying to
cram it into a platonistic/logicistic framework.

Joshua Stern

3.[urgent] natural language grammar and corpus

hello friends,

  I am student of M.Tech (or MS to say.. ) and have taken up course in
Natural Language Semantics and currently working on project
question-answering. For this purpose I am looking for a nice natural
lanuage grammar ( I don't want to go ahead of CFG or that level) can
anybody please help me in this regard....?????

and also for the same purpose I am looking for a nice corpus around
1000 lines or so which is quite good for my purpose... please
urgently, because deadline is approaching...



4.The language of aggrevation is Bush language

I hate it, no human should be allowed to be that
aggressive. So agressive that it feels like my
head is going to explode and that the world
will break.

5.Bush language and media language the same brutality (agressive business mentality)

Most may not agree with me on this. I know what I am
talking about. I am not a spammer. Right now I am
an anti-war basment philosopher, not an activist.

But there is a deep serious thought involved here.

6. [BVE PSA] Logical, natural language based programming language (t3d)

7. CFP: AIPR-07 Special Session on Frontiers of Language Processing and Information Retrieval for Asian Languages

8. is class-based language models the same as cache-based language models

Return to AI


Who is online

Users browsing this forum: No registered users and 78 guest