W3C home > Mailing lists > Public > semantic-web@w3.org > July 2007

Re: Terminology Question concerning Web Architecture and LinkedData

From: Tim Berners-Lee <timbl@w3.org>
Date: Tue, 31 Jul 2007 13:46:41 -0400
Message-Id: <CB2D3917-FAC7-4196-A4F3-915E9D6DDEE3@w3.org>
Cc: <conal.tuohy@vuw.ac.nz>, <semantic-web@w3.org>
To: <tim.glover@bt.com>


On 2007-07 -25, at 04:54, <tim.glover@bt.com> wrote:

>
>
>> the owner of a URI has the right to define what it identifies...
>
> But that right is unenforceable. McDonalds might define a "McJob" as a
> fulfilling career choice, but they cannot prevent other people  
> using it
> to mean something else.

You are confusing natural language and semantic web architecture.

This is true of "McJob" but it not true of http://mcdonalds.com/hr/ 
jobs#mcJob

They own that URI and can use it to

> Surely it is futile to insist that a particular string has only one
> meaning, ***EVEN IF it begins with the characters "http://" ***!!!

This  not a string, though.  When quoted, it is quoted a a URI.
The languages use URIs for symbols.

In N3 for exxample,   <http://mcdonalds.com/hr/jobs#mcJob> is a  
symbol owned by McDonalds (we imagine)
but "http://mcdonalds.com/hr/jobs#mcJob" is a string of characters.

In general strings out of context do
> In
> fact, maybe no two people ever mean exactly the same thing by a given
> string...

Strings out of context in human language have that property but URIS in
semantic web languages do not.


> (as an aside, I cannot help thinking that that a great deal of  
> confusion
> has been caused by using URIs for everything. Surely the meaning of a
> string cannot be determined by the fact that a piece of software leads
> it to another string - doesn't that lead to infinite regression..?

No, it does not. Again, you are taking across things from natural  
language.
If you recursively look up all the words in a dictionary definition,  
you will spend a long time.

What happens in the semantic web is that things bottom out in two ways.
When you look up a term, you load the ontology, which gives you  
(typically) OWL
statements about the term, and other things.  When you look up the  
predicates and types
used in that description, you find that (typically) the terms are  
taken from the OWL
ontology and also RDFS and Dublin Core,  so you can look up the terms  
one or two more
levels, but it bottoms out very quickly.

If you follow subjects and objects, all related things, then it will  
bring in huge amounts of data.
I believe the predicates and types, Properties and Classes, have a  
special role of defining the language being used.

Of course there is also in the ontologies a bunch of english text,  
often referring to other specifications, which help.
define, for the curious, the terms more carefully.  But that english  
text is the output of (typically) working groups, on behalf of the  
wide community of possible users,
who have iterated over ad nauseam, in a process which allows pubic  
comment, and includes the provision of
test cases to cover corner cases, and so on.  In this case, you can  
say "no two people ever mean exactly the same thing by a given  
string" but it becomes decreasingly relevant when the terms in use  
have gone through this process and then have been used for exchanging  
information between systems for years.  Importantly, there is often  
the possibility of opening up a new dialog when some uncertainty  
occurs, and asking the original creators, the owners.

> string

> Maybe
> it would simplify things to talk about strings of letters, and keep
> dereferencing as a special case. )
>
> Tim.

Tim

>
>
>
>
>
>
>
Received on Tuesday, 31 July 2007 17:46:55 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:58 UTC