W3C home > Mailing lists > Public > www-rdf-logic@w3.org > October 2000

Re: names, URIs and ontologies

From: pat hayes <phayes@ai.uwf.edu>
Date: Tue, 31 Oct 2000 14:58:26 -0600
Message-Id: <v0421010bb624d7f9482d@[205.160.76.86]>
To: Lynn Andrea Stein <las@ai.mit.edu>
Cc: www-rdf-logic@w3.org
Hi Lynn, thanks for quick response.

>Let me spell out what I think I understand, then see whether it's my
>understanding of logic, of URIs, or of your concerns that is broken.

The last, I think.

>LOGIC:
>
>A logical name is an arbitrary symbol that takes its meaning from the
>valid models of the theory (i.e., interpretations of that theory)
>within which the name is used.
>
>Two logical names spelled differently are different names.  They may
>have the same interpretation (in some or all models), but they are
>different names.  (In some logics, same-interpretation can be
>constrained through the use of an equality/equivalence operator.)
>
>A quantifier introduces a name whose spelling implicitly depends on
>that quantifier.  E.g., in [forall x . foo (x)] and [forall x
>. bar(x)], the x's are actually spelled differently because they're
>scoped by two different quantifiers.

Thats an odd way of putting it, but I see what you mean. If the scope 
were part of the spelling they would be spelled differently, yes.

>A logical name spelled the same way in two different occurrences
>refers to the same thing (i.e., has only one interpretation that
>covers both cases) provided both occurrences are within the scope of the same
>quantifier(s).

With the qualification noted, yes. But notice you are talking about 
logical 'names', ie constant symbols, not about proper names. 
Conventional logics don't contain proper names (by which I do not 
mean that a name like "Boston" can't be used as a logical constant, 
but that when it is so used it loses all the naming content it has as 
a word in English.)

>URIs:
>
>A URI is an arbitrary symbol that may be conventionally associated
>with an operational process by which further
>(non-logically-constraining) information or objects may be retrieved.
>This process is (often) useful in establishing social conventions.

Im not sure what that last sentence means, but OK.

>Note that a URI need not have such an associated process.

Ah, that is interesting. Can you give me an actual example  - a real 
one, from actual usage, not just made up to be a possible example - 
of a URI which does not have such an associated process? Is it 
possible to recognise a URI as being a URI just from its syntactic 
form, or can any name be used as a URI?

>There are three kinds of URIs:
>
>global URIs, which can be used anywhere and have the same associated
>operational process (and, when we treat URIs as logical names, the
>same intended interpretation) no matter where they are used.
>
>local-only URIs, which can be used anywhere but are in effect
>existentially quantified within the page of use (and for which any
>associated operational process is relative to the page of use).

Can be used anywhere? Not in DAML, though, right? What would it mean 
to use a local URI outside its scope?

>global-and-local URIs, i.e., local URIs with an explicitly spelled out
>global prefix.  These have two-step operational processes (if any) but
>logically behave like global URIs, i.e., they can appear anywhere and
>have the same intended interpretation (and operational processes) no
>matter where they appear.  To the extent that they are understood as
>(existentially) quantified names, the quantifier is the global prefix,
>NOT the page on which the URI appears.

These seem exactly like global URIs; there seems to be no logical 
reason to distinguish them (there may be pragmatic reasons, of 
course.)

>Two URIs with the same spelling are the same name (and have the same
>interpretation) iff the spelling includes a global part.  Two URIs
>with different spelling are different names, though nothing prevents
>them from having the same interpretation.

OK so far.

>PAT's CONCERNS:
>
>I think that Pat's worried that all URIs take meaning from the
>containing name.  But only local-only URIs do this, so I don't think
>it's an issue.

No, that is not my worry. My worry is that URIs are essentially 
logical constants with one of two kinds of scope: either very global 
or very local. And proper names (which are not logical constants in 
conventional logic) have a 'scope' (scare quotes because of previous 
note) which lies somewhere between these extremes, so a restriction 
to these extremes makes it impossible to use proper names; and that 
proper names are a vital tool in sharing knowledge, so to make them 
illegal is probably a serious mistake; and finally, that the 
development of a really useful web logic would indeed involve 
extending conventional logical methods of analysis to allow things 
more like real proper names, and they would be indeed like logical 
constants which are bound by existential quantifiers with a 'social' 
scope, something larger than a single source but smaller than the 
entire web. (If I were feeling poetic at this point I might start 
saying things about how it takes a village to properly define the 
meaning of a name....but maybe I'd better wait until after the 7th.)

Heres another way of putting the point. Imagine a heirarchy (tree) of 
places where one might want to 'anchor' a name, so that the name is 
understood to have the same meaning throughout the part of the 
heirarchy below the anchor. (Think of a parse tree of an expression, 
with quantifiers as the anchors.). Right now, this heirarchy has the 
entire web as its top node, and goes immediately to single web pages 
at the next step. It then, sometimes, goes even deeper. What I want 
is to have more layers in between the top and a single page. I want 
to be able to anchor a quantifier (namespace) somewhere in between, 
where it defines the shared usage of names by a web subcommunity. If 
the entire semantic web is like a huge logical expression, I want it 
to have a more useful syntax than just a world-wide bag of pages.

>I also think Pat worries that someone needs to pick the One True
>public name for Boston.  I think that this is the role of social
>convention, that this usage evolves,

I agree entirely, but the current view of DAML names seems to me to 
essentially make such social process impossible. I am stuck right 
now, unable to do my homework. I want to refer to the people who work 
here at IHMC, in DAML. How can I possibly do it? I can't use their 
names (well, I can, but then they no longer refer to those people, 
they are just existentially quantified variables on my DAML page), 
and I don't want to say that their web pages work for the State of 
Florida. How can I refer to them?

> and that there is already good
>precedent for this happening with, e.g., RDF M&S.

? I can't see any reference there to this issue. Can you point me to 
where it is discussed?

>I also think that
>we can provide suggested conventions (which the market will or won't
>adopt) and that we MUST provide some machinery by which two names
>spelled differently can be identified as synonyms (e.g.,
>daml:equivalentTo).  But again, this is analogous to the logical
>situation.

Yes, that is the logical strategy. But it is impractical to do this 
for every token of every name on the web.

Pat

---------------------------------------------------------------------
IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola,  FL 32501			(850)202 4440   fax
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes
Received on Tuesday, 31 October 2000 15:55:05 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:52:37 GMT