Re: AW: [ontolog-forum] Current Semantic Web Layer Cake

I am sure others here will have better answers that mine, but let me  
try...

On 6 Aug 2007, at 11:43, <tim.glover@bt.com> wrote:
> Henry wrote:
>
>>>>>>
>
> In logic programming you could always write something like
>
> Bachelor(x) :- Unamarried(x), Man(x) .
>
> But you knew that someone else could define those terms differently so
> you could not merge the work from different groups easily, even if you
> thought it was of the highest quality. The semantic web takes this
> global naming seriously, and makes it clear how all these technologies
> can work together, how they can be integrated. Neither UML, MOF,  
> nor SQL
> nor prolog nor any of the other languages made this clear.
>
> <<<<<
>
> Well I agree that the major problem of data integration is the fact  
> that
> people use the same words in subtly different ways. But can someone
> please explain to me, how does putting http://xyz in front of all the
> names help?

URIs, are Universal Resource Identifiers. They act the way names and  
predicates
are meant to work in mathematical logic.  But URLs have the following  
extra features:

- built into URLs is a notion of ownership. I own bblfish.net, so it  
is possible to
assign responsibility, one way or another to statements.
- It also allows distributed name creation. This is not easy when one  
uses mathematica logic.
Did someone elese somewhere use the P1 predicate? How do I know.  
Ownership helps me decide this.
If I coing names I own, I know I don't tread on other people's space.
- these URLs can be dereferenced

> "...someone else could define those terms differently..."
>
> If you mean by giving a different human interpretation of these words,
> that is beyond the scope of the semantic web. If you mean by writing
> alternative, inconsistent statements about bachelors and marriage, I
> have heard repeatedly that that is perfectly acceptable, if not
> positively encouraged!

yes. For example I could name the bachelor concept with http:// 
bblfish.net/phil#Bachelor
and you could do it at http://bt.com/~tglover/ont#Bachelor .

This is what analytic philosophers do all the time when they try to  
distinguish different
concepts, such as "knows". They will give a definition for say  
knows1, then knows2, etc...

>
> Is it because people who want to use a different definition can use a
> different URI? But how do they KNOW that their definition is the  
> same or
> different from someone else's?

This is where it helps to have dereferenceable URLs. By doing a GET  
on a word we can find
the core of its meaning.

> The difference in meaning lies precisely
> in the use people make of the words.

It is more complex than that. People can use words as Putnam pointed  
out, without knowing their meaning. I may be able to say that "Gill  
has Multiple Sclerosis" correctly, even if I know very little of the  
meaning of the word. The meaning is given by the specialists in the  
field, and by reality. I can use the word wrongly, that won't make me  
right, as Alice correctly told Humpty Dumpty.

Because dereferencing a URL makes it so easy to find a meaning, the  
owner of the URL, which can be a person or an institution, has a very  
large say on what it means.


> What is to stop me and a colleague
> across the planet from innocently and with the best of intentions  
> using
> the URI in different ways, each believing that what we are doing is
> correct in our own context?

Nothing. People can and will use words wrongly.

But where there is a right, there is a wrong. If we can dereference  
the URL, find the meaning, then that may help sort out the right from  
the wrong usage. Otherwise it may be that both of you are right. It  
would be interesting to come up with a real example here.


>   And having used different URIs, how do we
> integrate that part of our knowledge we agree on?

You can make statements such as

<http://bt.com/~tglover/ont#Bachelor> = <http://bblfish.net/ 
phil#Bachelor> .

or

<http://bt.com/~tglover/ont#Bachelor> rdfs:subClassOf <http:// 
bblfish.net/phil#Bachelor> .

or even perhaps with rules using SPARQL or N3 rules.


> Perhaps the answer is something to do with dereferencing?

yes.


> Well it was recently pointed out on this thread that a dictionary  
> is a mechanism for
> dereferencing ordinary names to content. But you seem to believe that
> this does not solve the problem for ordinary words.

The thing to notice here is that we are able to communicate. So  
insofar as you feel that the dictionary problem undermines real  
language, I would say it proves a limitation or misunderstanding of  
the dictionary argument.

In essence signs always send one to further signs. Yet we do  
understand each other. Questions have to end somewhere: I would  
suggest that they end in action. The W3C has put together a core  
vocabulary (RDF, RDFS, OWL, etc...) which will be well understood by  
interpreters in the field. People are building reasoners on these  
vocabularies. Those reasoners will do transformations immediately on  
perceiving those words. They will not go and try further dereferencing.


> Why is the web
> logically different from a very partial dictionary, where most of the
> entries are in Chinese?

Complete knowledge is impossible. When you browse the web you learn a  
lot, even if you can never completely own all of the web. The  
semantic web will work in a similar way. Different software agents  
will be satisfied by understanding different subsets of the space.

> Sorry to be so dense :(

I think it requires a small aha! moment to get it. I suggest playing  
with N3, downloading cwm, writing your own foaf file, and then things  
get to make a lot more sense. :-)

> Tim.
>
>

Received on Monday, 6 August 2007 10:36:00 UTC