[
Lists Home |
Date Index |
Thread Index
]
This is getting pretty philosophical.
I've never seen a case of discovery (in the sense of machines discovering
some important aspect of information they are receiving) that didn't boil
down to looking up the meaning a human being has already applied to this
information or this class of information. And it's also a human who
tagged the information such that the machine could even begin to figure
out where to look to "discover".
I'm not questioning the value of what we now refer to as discovery. I do
question any tie to the plain English meaning of the word "discovery".
Every single time a system returns results that are relevant, they are
relevant to what a human being said they should be relevant to, and they
are relevant in ways that the same or other human beings already defined
as meaningful.
It's the process of applying these criteria and then sifting through the
results, and redefining "relevancy" that gets faster and more fluid with
what we now call "discovery".
I don't think our understanding of either neural networks or of the human
brain has progressed far enough for any meaningful comparison to be made.
From what I've read (and I'm no expert) the two seem to be diverging.
As happens time and time again in philosophical conversations, two people
arguing about which of their views is "true" actually turn out to disagree
about which truths they are motivated to talk about. The fact the
language itself is a set of inadequately rigorous agreements to begin with
just adds to this problem.
--------->Nathan
On Thu, 13 Jan 2005 08:06:08 +1100, Rick Marshall <rjm@zenucom.com> wrote:
> i can't possibly put all my understanding of this problem succintly
> enough for a short email, but here are a few pointers to my ideas:
>
> my basic point is that owl et al are inadequate as a basic principle for
> semantic anything. why? well like roger alluded to in the english
> language comments - what makes evenly poorly composed english
> understandable? part of the answer is word roles - verbs are not just a
> syntactic device, they are semantically very different to nouns. if i
> simply give you the words "boy" and "ball" you have very little
> information. if i add to that word list a verb - "throws" you can start
> to derive meaning. in document terms "order" "goods" "buy" has
> meaning... "order" "goods" lacks purpose (unless you accept order as a
> verb). things like ebxml and uml try to address this and for my money an
> extension of this idea to be the fundamental of the semantic web is
> essential. the we can say things like "this message is buying goods -
> maybe it's an order".
>
> next there's the problem of recognition - i'm happy that a neural net is
> probably the only way to do this in a general sense. this implies
> training (constantly) - but then kids need school too and the brain is
> the most sophisticated and complex device we know. my view is that if
> the most complex and competent device in the known universe requires
> significant training to do things, then there's some sort of biblical
> arrogance in our assumption that we can simply write down some rules and
> build a device that works as well.
>
> i'd be very interested to know if anyone on this list has tried using
> neural nets rather than schemas to classify xml documents and/or if
> anyone is interested in setting up some experiments to try it out.
>
> and one final point - back to the sum is greater than the whole. i was
> thinking about this in terms of an element algebra. group theory defines
> a group by operations (verbs :) ) that when applied to members of the
> group (usually, but i guess not necessarily, 2 members - could be
> ternary operators) result in a member of the group. integer + integer =>
> integer. but if you have a group member you have no way of knowing if it
> was derived by operation (and there may be an infinite number of
> contruction operations), which one, or does it just exist in it's own
> right. the number 4 as an integer has different properties to the
> numbers 1 and 3, but can be constructed from them. if i have a bunch of
> logs or steel bars they could be the parts for a bridge or a scaffold or
> ladder etc. the construction again has some meaning beyond the parts.
>
> back to <>s. our scientific culture is deconstructive (break it down to
> find the meaning) while our engineering is constructive (put it together
> to make something) and we are creatures of action (use the things we
> make to do something) all the time fighting dS/dt > 0 (trying to create
> order from chaos).
>
> my answer to your question roger is in principle yes, but we need xml
> technologies in all 4 areas above to do it.
>
> rick
>
> Bullard, Claude L (Len) wrote:
>
>> And the fact of trading (that these components are created and
>> shared by some network of traders "dynamically assembled by
>> one system and shipped to another where the assembly is
>> dynamically understood") infers that some common maps already
>> exist, aka, the upper level enterprise/market ontology.
>>
>> Intentionality is the impetus of ontology.
>>
>> No matter how simple the intent is, it will shape the
>> understanding. It is a centrality and may be a transient
>> or permanent attractor that creates meaning.
>>
>> len
>>
>>
>> From: Bullard, Claude L (Len)
>>
>> From: Roger L. Costello [mailto:costello@mitre.org]
>>
>>
>>
>>> I assume that this question has as its impetus ...
>>>
>>>
>>
>> Here's my motivation for the question: in a large, complex Enterprise
>> you
>> may know the kinds of "things" that need to be moved around (e.g., Book,
>> BookCover, etc) but you don't have a-priori knowledge of the specific
>> transactions that will be needed.
>>
>> So, is it feasible to simply declare a bunch of components (that
>> everyone
>> understands), which may be dynamically assembled by one system and
>> shipped
>> to another system where the assembly is dynamically understood.
>>
>> -----------------------------------------------------------------
>> The xml-dev list is sponsored by XML.org <http://www.xml.org>, an
>> initiative of OASIS <http://www.oasis-open.org>
>>
>> The list archives are at http://lists.xml.org/archives/xml-dev/
>>
>> To subscribe or unsubscribe from this list use the subscription
>> manager: <http://www.oasis-open.org/mlmanage/index.php>
>>
>>
>> !DSPAM:41e58986150941861020313!
>>
>>
>>
>
--
.:||:._.:||:._.:||:._.:||:._.:||:._.:||:._.:||:._.:||:._.:||:._.:||:._.:||:.
Nathan Young
A: ncy1717
E: natyoung@cisco.com
|
