W3C home > Mailing lists > Public > semantic-web@w3.org > June 2005

Re: Stack of two towers

From: Jos de Bruijn <jos.debruijn@deri.org>
Date: Thu, 16 Jun 2005 19:25:07 +0200
Message-ID: <42B1B5F3.90306@deri.org>
To: Henry Story <henry.story@bblfish.net>
CC: "'SWIG'" <semantic-web@w3.org>

Hash: SHA1

Henry Story wrote:

> Hi, following some leads in the blogsophere I came across a paper
> entitles "Semantic Web Architecture: Stack of two Towers?"[1].
> I have a not had time to look at this in depth yet, and probably
> don't understand a few of the pieces of the puzzle. They are basing
> their argument on the difference between the open and closed world
> assumptions, and how this leads us to conclude different things
> from the same facts.
> [[
> Given an ontology containing only a single RDF triple:
> <#pat> <#knows> <#joe>
> the answer to a query asking if pat knows exactly one person would
> be “no” under
> RDF’s open world semantics, but “yes” under the closed world
> semantics of Datalog.
> ]]

This statement is very arbitrary and the answer really depends on the
kind of question you ask. In fact, here there seems to be a form of
closed-world semantics involved in the query, but not in the knowledge

Say, we have a knowledge base KB, which consists of:


If you query for all persons known to pat:

?- knows(pat,x)

You get the following answer:


You get this answer both under Datalog and under First-Order
semantics. You can afterwards count the number of answers and
interpret this number in any way you want. This is independent of the
logic used for the knowledge base.

If you want to query directly whether pat knows exactly one person you
need negation and then there is of course a distinction between
classical negation and negation-as-failure, but this is beyond Datalog.

> Now is this really a problem or is this one of these puzzles that
> can be unraveled by
> just noting a distinction?
> My thought on the following is that perhaps this is a problem that
> disappears as soon
> as one allows as N3 does, to speak of named graphs also known as
> labeled formulae - one
> needs not triples but 4-tuples.
> First to avoid ambiguity, let's change the example to something
> that does not itself involve
> named graphs. Knowledge, belief, desire, etc are all known in
> philosophy as propositional
> attitudes: ie they relate a knower to a statement or set of
> statements (named graphs). So to
> have an example that uses those terms, though perfectly reasonable,
> is just going to make
> it confusing to make the points that follow.
> So let me take the following statement that limits its elements to
> the world of objects
> (things that can be expressed clearly with triples and only triples)
> S: <#pen> <#is_in> <#pocket>
> Now the closed world assumption is a logic limiting itself to the
> deductions that follow from a graph, as if the graph contained all
> the facts in the world. So making deductions in a closed
> world always makes reference (sometimes implicity) to the graph in
> which the question was posed.
> We therefore don't really have one question which has two answers,
> one given a open world assumption, and one given a closed world
> assumption, rather we have two questions:
> OW) Is there more than one pen in the pocket?
> CW) Does the graph S contain information about there being more
> than one pen in the pocket?
> clearly each of these questions have different answers. I suppose
> to OW one would answer don't
> know, there could be more, whereas to CW one would answer no, there
> is no more information there.
> Both of these questions could clearly use the same ontology, one
> allowing more than one pen to be
> in a pocket.

Indeed. Both queries require a form of negation; the OW question
requires classical negation and the CW question requires
negation-as-failure (also called 'default negation').

Best, Jos

> Henry Story
> [1] http://morenews.blogspot.com/2005/06/two-semantic-webs.html

- --
Jos de Bruijn, http://www.uibk.ac.at/~c703239/
+43 512 507 6475         jos.debruijn@deri.org

DERI                      http://www.deri.org/
- ----------------------------------------------

Only two things are infinite, the universe and
human stupidity, and I'm not sure about the
    -- Albert Einstein
Version: GnuPG v1.2.5 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

Received on Thursday, 16 June 2005 17:25:24 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:46 UTC