Re: Comment on "Meaning and the Semantic Web"

From: Pat Hayes <phayes@ihmc.us>
Subject: Re: Comment on "Meaning and the Semantic Web"
Date: Fri, 4 Jun 2004 10:32:36 -0500

[...]

> The point I was intending to make was that the worries (about 
> allowing others to have hostage over your thoughts when writing an 
> ontology) which you and Peter expressed, seem to me to embody an 
> attitude towards the entire business of communicating on the Web 
> which, if translated directly into an attitude towards social 
> communication more generally, would appear as a form of paranoia 
> concerning any kind of communication: and someone who acted by those 
> maxims would finish up quite rapidly being completely cut off from 
> social intercourse of any kind involving language, and would probably 
> be happier as a Trappist monk under a vow of silence.  

Well, yes, guilty as charged.  If every term (not just common words, but
also proper names, noun phrases, etc.) I use commits me to information
created by the coiner of that term, then I would be happier under a vow of
silence.

Note, however, that this only holds for me if the commitment comes from
*every* *term* and from a place determined by the form of the term.  I am
perfectly willing to join communities that provide a common set of
information related to certain terms.  

> As I say, this 
> is what your concerns (vividly expressed in terms of totalitarianism 
> and control) seem to lead to; and, perhaps more personally, though it 
> was not intended to be ad hominem, the arguments and justifications 
> you, Bijan, give in response to objections seem to betray a similar 
> attitude, one that to me seems quite inconsistent with normal social 
> assumptions. You have said for example that in your view, when you 
> use a term introduced by someone else in an ontology you write, that 
> the meaning of the term should be entirely under your control, and 
> entirely determined by your ontology, with no reference whatever to 
> anything said by the 'owner' of the term. 

Yes, I want to retain this ability for certain cases.  I believe that most
terms will indeed have a commonly-used meaning, and that this meaning will
usually be provided by the coiner of the term.  I fully realize that there
is a cost to be paid by not committing to this meaning, and that this cost
can indeed be quite serious.

> I find this an 
> extraordinary claim, and  intended to make this sense of unreality 
> vivid by drawing from it what seems to me to be the obvious 
> conclusion: that such an attitude, applied to the entire social 
> fabric of a Web, implies that writers of ontologies are only 
> concerned with being right, and not with communication at all. For 
> presumably you allow others the same leeway that you claim: so if 
> they use these same URIs in their ontologies then their meanings need 
> have no relation to your meanings; so we are left with a picture in 
> which all ontologies are entirely private worlds, whose use of common 
> URIs is merely a syntactic accident. 

Yes, this could happen in our scheme.  However we do allow that different
users of a term can commit to the same meaning of that term, either by
independently incoporating equivalent (or similar) information related to
that term, or, preferably, by both importing a document that provides
information about that term.  This imported document can easily be the
standard document provided by the coiner of the term.  We thus do not
expect that communication will collapse because of differing definitions
for most or all terms.

> And this is good, presumably, on 
> this view, because any kind of presumed relationship between 
> ontologies would render everyone's meanings hostage to others' 
> misinterpretations: a danger - if you wish to think of it as a danger 
> - to which of course ANY form of communication is prone; so if 
> avoiding this danger is your primary motivating concern, then it 
> seems to follow that you are advocating that communication of content 
> is not the primary aim or goal of the Semantic Web. 

Well, yes, this is a danger.  I believe that it is not a negligible danger
- for some terms - and thus that a communication scheme must have ways to
opt out of information provided by the coiner of a term.  I believe that
human communication works this way - usually we go along with the standard
information package related to meaning of a term, e.g., the president of the
United States, but sometimes, and sometimes even for good reasons, we opt
out of this standard package, perhaps to argue that the true president of
the United States is not the person annointed by the Supreme Court but is
instead the person who should have received the most electoral votes under
the rules in place at the time the election was held.

> So my riposte, 
> which gave you offense for which, to repeat, I apologize, was 
> intended to be along the lines of a standard response one can make to 
> someone who advocates solipsism: if you are right, then I don't 
> exist, so please stop talking to me about it.

Well I am (currently) prepared to believe that you exist and, moreover, to
commit to the information about you that is contained in your home page.

> No doubt I have this conclusion wrong, or it is not what you 
> intended. However, as I say, to me it seems the only conclusion that 
> is consistent with the position you advocate and the reasons you give 
> for advocating it. I would be delighted to have the error in my 
> reasoning explained.

Well, to summarize, I view the opt-out aspects of our proposal to be there
for the hopefully rare cases where they are needed and that communities, in
particular application communities, will in fact end up always using the
coiner-provided meaning of the terms related to their interests.  

[...]

> Pat

Peter F. Patel-Schneider
Bell Labs Research

Received on Friday, 4 June 2004 14:59:19 UTC