RE : semantics and knowledge

Hi Howard,

I won't argue much with that. The problems that you are mentioning are acknowledged as important by the SW community, and indeed have been part of the semantic web plan for a while
(http://www.w3.org/2007/03/layerCake-small.png : "trust" and "proof" are the items of interest for you there)
But SKOS is certainly not about tackling all these issues. Indeed we hope that solutions will come one day, but at a more general level. The problems you mention apply to any kind of RDF-encoded "knowledge"...

Antoine

-------- Message d'origine--------
De: public-esw-thes-request@w3.org de la part de Howard Burrows
Date: jeu. 06/11/2008 15:57
À: public-esw-thes@w3.org
Objet : semantics and knowledge
 
All,

 

I guess I don't feel that we get to talk about "knowledge" at all if we are
limited to semantics alone.  As Antoine indicates, this project is Semantic
Web-biased.  In that context, we don't get "simple knowledge"--we don't even
enter the realm of knowledge.

 

Semantics allows us to agree on word usage and document/data formats, so we
can explicitly describe things to each other across communities-so we can
clearly elaborate distinctions between our beliefs and hypotheses.  Doesn't
"knowledge" require more than that--actually justifying which of these
beliefs and hypotheses are right and actionable?  Aren't the requirements
for knowledge beyond what we're doing here?

 

As an example, I'm hoping my concern with the use of the word "knowledge" is
*not* just semantics; I hope I'm raising a legitimate issue, and not just
drawing in a term from an ontology parallel and possibly incommensurable
with the one used here.  This is important as the next phase of web
development should go beyond just understanding how words and data relate to
each other.  We need to move into the knowledge domain and establish
standards of a different kind that allow us to catalog statements, not just
according to what we understand them to mean, but by whether or not they are
true (or at least sufficiently justified to entitle us to use them in making
decisions).

 

As another example, Aida's hadron scheme states quite clearly a hypothesis
about the possible relations between agreed upon objects.  It doesn't get at
whether this is the correct hypothesis--or how we would tell.  What
observations should count as data in establishing the truth of this
hypothesis?  Even if we have ways to justify this particular hadron scheme,
we still need to be able to locate competing hypotheses to compare with this
scheme.  And we need to be able to rank the whole set of ideas that people
have come up with according to how well each is justified.  

 

All this seems somewhat beyond "semantics".  I'm still not getting the
perspective that most of you seem to share that would entitle us to use the
word "knowledge" in the name of this valuable semantic web standard.  Just
organizing the list of hypotheses (both right and wrong) and data (both
relevant and irrelevant) doesn't capture the critical relation between them
that generates knowledge.  Even if you prefilter the set of hypotheses so
you only include those hypotheses that qualify as "known-to-be-true", the
correct statements are not the knowledge-what justifies them is--otherwise
you haven't captured essential elements that would allow you to apply them
as knowledge.

 


Howard Burrows

Supporting Research

Durham, NH, USA