xml-dev - RE: Semantic Web "Definition"

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

From: Sean B. Palmer [mailto:sean@mysterylights.com]

>> Thus far the semantic web is the application of semantic
>> network technologies to the problems of discovery and
>> association of resources.

>That is only a very small part of a rapidly growing community: RDF/RDF
>Schemas, ontologies, logic systems, network databases, meta-information of
>all kinds, annotation systems, RDF API's...and so on. Semantic network
>technologies are almost the last part (SW specific API's are still a dream)
>of the development.

Do you have a definition or named relationship which could be used to 
identify this community as distinct from other communities?  How 
would a member of this community recognize another member without 
exchanging a badge?   What I am afraid of here is that Semantic 
Web is justAnotherBadge for a community of researchers who need 
approval for their research and therefore must cast it by a name 
which will raise the probability of approval:  riding the wave 
of one man's popularity without regard to creating requirements 
for cooperation (eg, one that let's one know if the process 
closes successfully or simply changed state).

>> Semantic networks evolved
>> from late sixties research into case grammars.  This is
>> not ambiguous or fuzzy.  It describes a meta-data
>> service to which other services can be routed to
>> aid in classification and pattern recognition tasks
>> involving text-based conceptual descriptions.

>I agree, but this is mainly semantic with a small "s", not Semantic Web,
>rather semantic as in meta-data classifcation techniques etc.

I would consider this S the big S in the Boltzman equations 
refined by Shannon and Wiener.  As such, given some value 
for statistical entropy, the semantic should enable me 
to narrow the "choices", ie, in Shannon's terms a means by which  
information is *assigned* meaning by choosing among equals.   
So far, you as the man you cite, still fail to assign meaning to 
the term "Semantic Web".  That is almost funny if it were 
not taken seriously by people who should know better.

>> The debate on the meaning of the term is intense
>> given the importance and source of such assertions as
>> "the semantic web is the future of the www[...] Requirements
>> such as "describing every possible aspect of your data"
>> are impossible to meet and will not be considered
>> seriously

>You think? That is the "aim" of the SW: where everything is so connected
>that proof validation becomes possible. Describing every possible aspect of
>your data will become possible (athough according to TimBL, proof
generation
>will not be a required part of the SW).

"We don't do tests."  the Wizard's helper in DragonSlayer.

You can achieve such connectedness in carefully limited 
domains.  So, this technology is a set of controls for 
limiting domains for closed systems (non-dissipative).  
Otherwise, you will find (see Szilard et al) that you 
need information to control information and that 
reasserts the primacy of irreversibility (it works 
once in one direction).  The need for information grows 
and therefore the system always exhibits uncertainty.

>> The nebulousness of the term "semantic web" leaves a
>> lot to be desired if one is trying to write hard
>> requirements for scheduled projects.   A precise
>> description of a meta-data service based on sharing
>> RDF documents has a better chance of being taken
>> seriously.

>Are you saying the SW isn't being taken seriously? If so, I could cite all
>RDF/DC/W3C meta-data efforts as proof to the contrary! 

I can cite a hundred societies based on devil worship.  That is not a 
requirement or evidence that I or anyone else should take the 
devil too seriously.  However, I do consider the efforts of 
communities of interest to commit to classification projects 
worthy.  After all, the markup community and hundreds of other 
language-driven communities do that every day.  What I do not 
take seriously is that this is a concept any different from 
those being undertaken in these communities or something 
that can be traced to the writings of the web seniors as revelatory. 
In effect, they are doing what all good taxonomists do and 
have done for a long time.   As Shannon wrote, (paraphrasing), 
semantics should not get in the way of engineering.

>Could you give an example of a meta-data based service for me? I agree that
>meta-data based services will be easier to explain/describe etc. than the
SW
>as an entity in itself; imagine trying to concisely explain the WWW. 

Let's do this by response to a problem that you can then summarize in a 
requirement.  Someone just sent me this URL:

http://www.lucifer.com/virus/virus.96/

If I try to follow this, a filtering service used by our company 
returns

"SmartFilter denied access to the URL http://www.lucifer.com.
----------------------------------------------------------------------------
----
It matches the category Cult and occult. "

Dang devil worshippers. ;-)  It seems someone has imposed a belief 
system on a URL.

Is it right?  Can I know this without reading the material?  How can 
I know this without disabling the filter?  How does a semantic 
network help here given that by reading that answer "matches the category", 
I might infer one has been used.  Probably a weak one (QoS issues).

Should I trust it or is trust simply a matter of policy with regards 
to the reliability of the filtering service?

Somewhere, somebody imposed a belief system on my ability to 
access a resource.  So would the semantic web protect against 
that or make it worse?  Only people shoot people, i know, but 
you seem to infer this can all be done by machines if only 
we accept the correctness of the Semantic with a Big S.

>That is in essence what we are trying to do: the SW is evolving from the
WWW!

Semantic networks evolved from research into case grammars.  Applying a 
technology to create a reliable service is not evolution.  It is
engineering. 
If by using some aggregate of such services, a new feature emerges not 
known before and which enables the system of services to adapt to a 
change in the environment or dominate a facet of a niche of rewards, 
that may be evolution.  

Perhaps it is good to discriminate learning from evolution based on 
the principle of irreversibility.  What does it mean to unlearn or 
to devolve?

That's a trap, BTW.  Go into it and you will find that semantics are 
largely like politics; local.  That is why semantic nets don't scale 
very well and why predicating the future of the internet on them 
is *dicey*.  Chance and the river...

A contractural service on the other hand does tests.

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h