Reto Bachmann-Gmür on Thu, 20 Dec 2007 04:25:23 +0100 (CET)


[Date Prev] [Date Next] [Thread Prev] [Thread Next] [Date Index] [Thread Index]

Re: <nettime> Critique of the "Semantic Web"


Florian Cramer wrote:

> On Wednesday, December 19 2007, 10:40 (+0100), Reto Bachmann-Gm?r wrote:
>   
>>>    The Semantic Web promises to overcome folksonomies with one, unified
>>>    and standardized keyword tagging system that can applied to anything.
>>>       
>> This is the wrong assumption that makes the further critique of the
>> Semantic Web missing the point. The semantic web is not about building
>> one big upper-ontology but about supporting many (usually very small)
>> ontologies. 
>
> But the design involves one-to-one mappings between these "ontologies" to
> ensure the correctness of a meta tag across them, so that for example a
> tag "Nettime" would always properly refer to this mailing list instead
> to the homonymous English company that produces software for tele caf?s.

In the technologies underlying the Semantic Web[1] "Nettime" neither
refers to a mailing-list nor to a company. "Nettime" is a literal and as
such self-denoting[2].  Referring to the mailing list is possible by
indirection using an ontology, the following says "There is something
that is a mailing list and is named 'Nettime'" (using an example
ontology ex as well as the rdf and dc vocabularies, using N3 syntax, the
prefix "rdf:", "ex:" and "dc:" are shortcuts for URI-namespaces):

[ rdf:type ex:MailingList; dc:Title "Nettime"].

This doesn't contradicts the following:

[ rdf:type ex:TeleCafeCompany; dc:Title "Nettime"].

which says "There is something that is a TeleCafeCompany and is called
'Nettime'".


The conflicting meaning you talk about is possible for URIs, one source
might say

ex:Nettime rdf:type ex:MailingList.

and another

ex:Nettime rdf:type ex:TeleCafeCompany.

If we trust both sources and assume the classes ex:MailingList and
ex:TeleCafeCompane to be disjoint we have a contradiction, the semantic
web doesn't prevent us from having contradictory knowledge, but under
some circumstances it allows computers to identify the contradiction.

There is not system to ensure the correctness of meta-data and the
decentralized semantic web could impossibly be. The semantic web
technologies allow to express arbitrary graph structures rather than
rigid structures as in traditional databases or hierarchies as in XML.
Graphs are better suited to match data from different sources: a partial
overlapping might be enough for a connected union graph, both of XML and
traditional databases unions are only possible if the sources to be
merged share a common schema.

Another aspect of the Semantic Web is the open world assumption, this is
the basic assumption determining the reasoning possibilities that we
never have all information about a universe. There are no default values
and we don't assume anything to be false (or true) just because we don't
know anything about it. Adding information to our model (graph) can
reduce the number of possible worlds in which our model is true but will
never make the model true for a possible world in which it was false
before. Again, this is no guarantee for the quality/correctness of the
data, but a good foundation to deal with multiple models of reality.

> But with these cross-mappings, you effectively have one tree on a meta
> level. 

Here I don't understand you, ontology alignment can make the merged
knowledge more concise and thus more usable, but where's the tree? And
why just one? With or without cross-mapping you have an arbitrary
numbers of graphs.

Reto


1. The term "Semantic Web" refers to an activity of the W3C and is
associated to a set of standards (RDF, RDFS, OWL, ...)
http://www.w3.org/2001/sw/.

2. See http://www.w3.org/TR/rdf-concepts/#section-Literals


#  distributed via <nettime>: no commercial use without permission
#  <nettime>  is a moderated mailing list for net criticism,
#  collaborative text filtering and cultural politics of the nets
#  more info: http://mail.kein.org/mailman/listinfo/nettime-l
#  archive: http://www.nettime.org contact: nettime@kein.org