Selected Newsgroup Message
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: question
Date: 02 Jun 1999 00:00:00 GMT
Message-ID: <37553090@news3.us.ibm.net>
References: <928286184.23137@www.remarq.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 2 Jun 1999 13:24:32 GMT, 166.72.21.169
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
mjcarls1@students.wisc.edu wrote in message
<928286184.23137@www.remarq.com>...
>hello. my name is michael carlson. i'm an undergraduate
>from the University of Wisconsin, Madison, with a degree in
>linguistics. i don't know much about computers yet, but i
>hope my question makes enough sense to be answerable.
>
>an assumption of linguistics is that human language is a
>biological system which provides some adaptative advantage
>for human survival. that language offers some survival
>advantage to humans (i don't presume to define that
>advantage here. rather, i am using 'advantage' broadly.)
>entails that previous to the emergence of language, humans
>existed in some biological state x which was, unless
>rectified by the biological emergence of language, in
>threat of danger by the external environment. therefore,
>like other biological systems (such as the circulatory or
>respitory systems), language is non-trivial in human
>survival.
>
Hi, Michael,
Language as an evolutionary advantage is the main thesis
of Steven Pinker's "The Language Instinct" book, which is
probably among the best treatments of the subject. But I
have some doubts about his conclusions, mostly because
of some indications that evolution is slow and language
evolution is fast and often because the latter can be
seen as emerging from dynamic environments with several
competing agents. You may take a look at the thread
"language as a result of natural selection" in this
newsgroup for more ideas about this.
> given, then, that some essential, biological relation
>exists, between human language and human survival, we might
>ask what place a natural language system would play in the
>'mind' and make-up of a computer. can we reasonably expect
>to program a computer with a natural language when the
>biological and environmental conditions needed for the
>emergence of language are not satisfied?
>
I may agree with you that we have a good problem when trying
to make computers understand and generate natural language,
but I wouldn't say for the reason you mentioned. The
environmental conditions that a computer (or, more adequately,
a robot) would find today can be seen as very close to
the conditions a baby finds when it is born.
The problem is not the environment, but the computer's "vision"
of this environment. Several of today's natural language
programs don't find a place for "knowledge of the world" *prior*
to the learning of language. They assume that natural language
is the result of syntactical competence, when in fact it
appears to demand *previous comprehension* of the world. Let
me clarify this a bit.
I propose here that you cannot understand language unless
you understand what language is talking about (the subject
complicates a bit with adults where the process is often
reversed, but I'll let this subject to another message).
A baby starts to learn words only *after* she is able to know
what the word refers to. Comprehension precedes language,
and this thesis allows us to think about the *groundings*
that syntactic constructs must have.
In other words, disembodied computers may have difficulties
(to say the least!) to "understand" language, because of
the lack of support for the concepts that lie behind syntax.
>alternatively, we might ask, Is natural language the same
>as human language?
>
No, it is not. Human language (on a more general aspect)
includes body language (facial expressions, hand movements,
body posture, eye movements, etc) and also things like
intonation, stress and speed. It is not easy to imagine robots
doing these things naturally. The thing is very complex.
>further, it seems like if we define the biological
>condition satisfied by language, we can construct an
>artifical language, but not until then. for example, unless
>we understand the function or biological role of the lungs,
>we cannot construct a mechanical, artifical lung.
>
I think this can be said of the brain, not the language.
This is another good reason for all AI "aficionados" to
study Cognitive Science and Neuroscience. The solution
to NLP will be found there first.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: question
Date: 02 Jun 1999 00:00:00 GMT
Message-ID: <3755aa23@news3.us.ibm.net>
References: <928286184.23137@www.remarq.com> <37553090@news3.us.ibm.net>
<928336205.26260@www.remarq.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 2 Jun 1999 22:03:15 GMT, 129.37.182.115
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
mjcarls1@students.wisc.edu wrote in message
<928336205.26260@www.remarq.com>...
>Dear Sergio,
>
>thank you for your response.
>
>i was particularly interested in what you said about wrold-
>comprehension grounding linguistic development. for all
>the time i've studied linguistics here, it's seemed pretty
>clear to me that without the perceptual faculty, language
>cannot exist. for example, we might imagine someone born
>(God forbid) without any means of perception at all: it
>seems rather implausible to me that the concept of 'noun'
>somehow exists in that individual's mind when they have no
>perceptual referent which gives rise to the concept 'noun.'
>perhaps chomsky is wrong, then, in claiming that linguistic
>structures are neccesarily innate. at the very least, our
>non-perceptual speaker shows that such an idea cannot be
>tested for.
>
One of the interesting cases to explore in this regard is
that of Helen Keller. Helen became blind *and* deaf at age
2, which means, she practically lived all her life with
just touch (taste and olfaction don't help very much).
She was able to develop language and as adult she
"wrote" a book. But one can't say that she had innate
grammar to help in her task of learning language. Touch
is a pretty powerful sense, and that allowed her to
*understand* some basic concepts in which to ground
language (such as the perception of the limits of her
body, the perception of someone else's body, the concept
of property, her hand, someone else's hand, her toy, etc).
It is from these notions that she probably built her
cognition, which resulted in the development of language.
> i wonder this,though: given that a lexicon is
>perceptually derived, it makes sense to me to go a step
>further and state that the syntax is perceptually derived
>as well. if an utterence is a 2-dimensional (linear)
>representation of a 4-dimensional perception, then we can
>ask two questions: 1) can a 2-dimensional symbolic
>representation of reality capture the essence of a 4-
>dimensional perception; 2) how would the linguistic system
>go about doing this?
> in answer to 1), i would contend that three dimensional
>(SPACE) perception is captured in the grammatical category
>of NOUN, and the interaction of SPACEes through TIME is
>captured in the grammatical category of of VERB. these are
>the root categories from which the grammar is dervied: the
>others (such as ADJ and DET) are just perceptual and
>conceptual expansions upon these two perceptual
>fundementals.
This is an interesting view and I find some correlations with
some neuroscientific findings. Verbs, which are usually linked
to *actions*, are commonly processed in a different area of
the brain than nouns, which are linked to actors (agents).
> as to how the system would arange this symbolic
>reduction of the perceptions, it seems as though the nature
>of human CAUSE - EFFECT perception (already a perceptual
>reduction in itself) does this for us. a very simple
>scheme:
>
>
SPACE
TIME SPACE
>
CAUSE EFFECT
> SUBJ
INT OBJ
> N
V
N
> JOHN
KISSED SALLY
>
Although this model appears to be a simplification of what
usually happens, I believe that it follows a close relationship
with the way children perceive deep semantic relations, which
can later surface in syntactic constraints, that "model"
grammar. Cause and effect also affects the creation of
a "world model", that will later be used to reason
abstractly and creatively.
> in a somewhat unrelated note: we were always taught in
>my syntax and semantics classes that truth value obtains at
>only the sentential level. this doesn't seem right. it
>seems more correct to state the truth value takes place at
>the lexical, phrasal, and sentential level. i.e.,truth
>value is hierarchical.
Yes, and there's also a pragmatic level, where the contextual
aspect is solved. I find each level of this hierarchy as being
constituted of characteristic patterns of usage that appears
to be learned all at once when a child listens to phrases.
Doubts in one level may be solved by clues obtained by the
analysis in other levels.
> for example, imagine i am looking
>at a particular something on the table. i lexically
>identify this as a pen. now, if language functions as a
>truth-generating feedback loop between perception and
>cognitive symbol making, then my lexicon continues to
>establish the validity of this perception, insofar as it
>continues to identify the pen as a pen. if, upon further
>inspection, i find that i have misidentified the obnject of
>my perception as a pen, when it was in fact a feather, then
>the feedback loop between my perception and language
>establishes this by labeling this object of my perception a
>'feather.' this process can be extrapolated for all
>hierarchical levels of sentential construction.
>
> anyways, i hope any of this is relevent. if there's
>something worth replying to here, i look forward to hearing
>your thoughts.
>
My idea of this loop you suggest is a bit different. I
follow the "perceptual" cognitivists. I put a perceptual
mechanism between "us" (our awareness) and the world.
This perceptual mechanism is "fed" primarily from the
stimuli that we receive from the world. This mechanism
is refined progressively, as the number of our experiences
in the world accumulate. But perception is also affected
by more inner levels (after perception).
Imagine that you see in a table a salad bowl. Your
perception tells you that it is a bowl. However, if you're
later informed that you're in a bar where a contest for
the biggest beer drinker will happen, that bowl may
start to look like a giant cup.
That's not the best example in the world, but I guess
you can get my point.
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: question
Date: 03 Jun 1999 00:00:00 GMT
Message-ID: <37569e5f@news3.us.ibm.net>
References: <928286184.23137@www.remarq.com> <37553090@news3.us.ibm.net>
<928336205.26260@www.remarq.com> <3755aa23@news3.us.ibm.net>
<7j5tvt$cjs$1@quine.mathcs.duq.edu>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 3 Jun 1999 15:25:19 GMT, 200.229.240.8
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
Patrick Juola wrote in message <7j5tvt$cjs$1@quine.mathcs.duq.edu>...
>In article <3755aa23@news3.us.ibm.net>, Sergio Navega <snavega@ibm.net>
wrote:
>>
>>One of the interesting cases to explore in this regard is
>>that of Helen Keller. Helen became blind *and* deaf at age
>>2, which means, she practically lived all her life with
>>just touch (taste and olfaction don't help very much).
>>
>>She was able to develop language and as adult she
>>"wrote" a book. But one can't say that she had innate
>>grammar to help in her task of learning language. Touch
>>is a pretty powerful sense, and that allowed her to
>>*understand* some basic concepts in which to ground
>>language (such as the perception of the limits of her
>>body, the perception of someone else's body, the concept
>>of property, her hand, someone else's hand, her toy, etc).
>>It is from these notions that she probably built her
>>cognition, which resulted in the development of language.
>
>Ahem. If she became blind and deaf at the age of 2, then I suspect
>she already had a hell of a lot of language development by then.
>
Well,...yes. By age 2 she would know a hell of a lot of things
about the world to support language. So we really can't conclude
that language, in her case, emerged only by the sense of touch,
but that she was able to *improve* her linguistic abilities based
only on touch. It is not all we wanted but it is something...
Regards,
Sergio Navega.
From: "Sergio Navega" <snavega@ibm.net>
Subject: Re: question
Date: 03 Jun 1999 00:00:00 GMT
Message-ID: <37569e65@news3.us.ibm.net>
References: <928286184.23137@www.remarq.com> <37553090@news3.us.ibm.net>
<928336205.26260@www.remarq.com> <3755aa23@news3.us.ibm.net>
<928379565.594@www.remarq.com>
X-Notice: should be reported to postmaster@ibm.net
X-MimeOLE: Produced By Microsoft MimeOLE V4.71.1712.3
X-Complaints-To: postmaster@ibm.net
X-Trace: 3 Jun 1999 15:25:25 GMT, 200.229.240.8
Organization: Intelliwise Research and Training
Newsgroups: comp.ai.nat-lang
mjcarls1@students.wisc.edu wrote in message
<928379565.594@www.remarq.com>...
>mr Navega,
>
>(i wish i could repsond to your post in particular, with
>quotes. apparently my news server (supernews.com) won't
>allow me to do this.)
>
>i think the example you present about Helen Keller suggests
>that _some_ perceptual faculty is required which generates
>the possibilities of syntactic structures. perhaps the
>lingusitic syntax derives from the 'perceptual syntax,' if
>we can define what a perceptual syntax would look like.
>that is, we _perceive_ the grammar; or rather, we perceive
>in terms of the grammar. in Keller's case, she could
>perceive that objects of the perceptual world have SPACE,
>and interact with one another through TIME. touch could
>reveal the perceptual syntax in this case.
>
We can say that some kind of sensory perception is necessary
in order to develop deep semantics in which to associate
syntax. This means that a brain which is fed only with syntax
(say, a brain linked only to a keyboard, if that was
possible) would be able to understand the syntactic level
and develop reasonable grammars, but not the "meaning" (by
the way, this is related to the chinese room thought experiment
of John Searle). To acquire meaning, the brain would have
to use some kind of sensory experience in order to ground
its syntactic structures.
>
>it seems like our discussion thus far is slowly redefining
>the linguistic system. rather then thinking about language
>as dervied through a 'Language Acquistion Device,' which
>effectively sees language qua language as an end in itself.
>however, might we be ablet to define language as one tool in
>a more broadly defined 'Knowledge Acquisition Device,' which
>somehow maps the lingusitic grammar onto the perceptual
>grammar in order to generate meaningful symbolic output? a
>'meaningful' utterence under this scheme would somehow
>satisfactorily map the symbols onto the percept without any
>lingering or un-mapable linguistic symbols.
This appears to be meaningful to our discussion so far.
In particular, LADs are things that are concerned
with the acquisition of syntax only, but I think these
mechanisms can't survive if "left by themselves". There
is some kind of "cooperation" with deeper levels that
are built throughout our sensorimotor interactions with
the world. This is, in my opinion, the kind of thing that
the proponents of the "poverty of stimulus" fail to
notice.
Then we can really talk of a Knowledge Acquisition Device
which will include not only LADs but also the kind of
knowledge that we capture from the world via our senses.
But it is not all there is to it: there is also another
way to acquire knowledge, and this is through thinking.
This is close to one of the things that I'm carefully studying:
that our knowledge of the world, which in children are
fed only by physical interaction, in adults is also fed
by language. It goes like this way:
a) Babies are born with a "fully open throtle", capturing
whatever falls in their senses. They start perceiving
regularities in phonemes, they start exercizing babbling,
they start to see how to control their limbs and a lot
of other things.
b) After some months, words start to be recognized and
associated to specific sensory perceptions (the ball, the
toy, mommy's face, some action verbs, etc).
c) Some time later, concepts of "ownership" develops. The
child recognizes her arms, her friend's toy, etc. Some
emotional conflicts start to appear (she wants her friend's
toy). This phase will be (according to my vision) the
groundwork where future highly abstract concepts will stand
(justice, property, transferrence of ownership, etc).
d) The child is now a minimally competent linguist. Her
sensory learning will continue as she discover new
things (like the fluididy of water, the colorful
reflex of light on transparent things, etc). But there
is a large body of "knowledge" about the world already
in place (gravity, inertia of objects, permanence of
objects, solidity, textures, colors, shapes, pain
provokers, etc, etc).
e) As the child goes to school, she will start to learn
more things through language and less through direct
contact with new sensory experiences. Part of her
knowledge *begins to be fed by language alone*!!
Obviously, there's a strong influence of visual things,
like a map of Egypt showing its location in Africa. But
this can be considered a *form of language*, as the
map is only a symbol standing for the "real thing".
This is the kind of mapping that other animals can't do
effectively. And guess what: this is related to another
important aspect of our intelligence, analogy and metaphor.
f) As adults, we have a lot of sensory knowledge of
our world (we still learn a lot, for example, as
when we take lessons to play golf or when we buy a
new car). But there is a huge number of "symbols" in
our head and we learn *a lot* from the "syntax" of
these symbols, through language, visual icons, maps,
drawings, etc. Reading a book, for instance, is mostly a
different kind of "sensory experience", one in which
we transform the syntax of what we read into a lively
collection of semantic (and sensory) patterns that we
had developed since childhood. In other words, reading
a book is mostly the act of obtaining *new* knowledge
by the use of *novel groupings* of sensory patterns
that we already had! It is mostly a "copy and arrange"
process, which makes a book look like a "recipe" of
how to assemble a new toy with the Lego parts that
are inside our mind.
g) But we have gone even farther. A mathematician may
spend his/her days looking to the scribblings in a piece
of paper and thinking, assembling "virtual worlds" in
his/her mind. The result is the development of new
constructions that don't have a physical counterpart
and that may help in one's task of creating absolutely
new things. Which is usually known as creativity.
Regards,
Sergio Navega.
Back to Menu of Messages Sergio's Homepage
Any Comments? snavega@attglobal.net