Re: "In Defense of Ambiguity"

I think it all boils down to the question whether we use URIs as "words" 
or as "identifiers" on the Semantic Web. Most will likely say that 
clearly, they are identifiers. However, it is pretty difficult to define 
and keep consensual the meaning of identifiers, because how shall we 
develop and renew our shared understanding of an identifier's meaning 
other than by using it and probing our understanding until it breaks 
every now and then? And then we are basically using URIs as words, not 
as identifiers.

One could hope that just using a higher level of abstraction will 
mitigate the problem - the more general we are, the more likely it is 
that subtle differences in meaning don't get relevant. However, almost 
all problems that we are trying to solve by semantic technologies are 
about handling extremely specific conceptual entities. We don't search 
for "agents" on the Web, but for "Java programmers with 3+years 
experience in Jena, and fluent in French and Mandarin".

Best

Martin

Richard H. McCullough wrote:
> Let's be explicit.
> Any thoughts of "negotiating" or "refining" the meaning of a term --
> we're talking about the "context" in which the term is defined.
>
> RDF/OWL people have little experience with "context" -- to them
> it's basically a namespace.  RDF/OWL really doesn't have "context"
> in its vocabulary.
>
> OpenCyc explicitly addresses "context", which is referred to as
> a "microtheory" (forgetting space,time for the moment).
> OpenCyc, in its attempt to capture common-sense knowledge,
> has defined  thousands of microtheories.  They are facing the "context"
> issue head on, and are making some progress.
>
> Until RDF/OWL introduces the concept of "context",
> in a form similar to CycL's "microtheory', or mKR's "view",
> you won't make any progress in this area. .
>
> Dick McCullough
> http://mKRmKE.org/
> Ayn Rand do speak od mKR done;
> knowledge := man do identify od existent done;
> knowledge haspart proposition list;
> mKE do enhance od Real Intelligence done;
>
> ----- Original Message ----- From: "Martin Hepp" <martin.hepp@uibk.ac.at>
> To: "Alan Ruttenberg" <alanruttenberg@gmail.com>
> Cc: "Booth, David (HP Software - Boston)" <dbooth@hp.com>; "Ian 
> Emmons" <iemmons@bbn.com>; <semantic-web@w3c.org>
> Sent: Tuesday, July 15, 2008 11:53 PM
> Subject: Re: "In Defense of Ambiguity"
>
>
> Hi Alan:
> Basically all I wanted to say is that in human communication, we clarify
> and refine the meaning associated to words in the course of
> communication, while the current SW infrastructure requires us to define
> the meaning of a conceptual element identified by a URI beforehand.
> Quite clearly, there can be multiple similar elements with different
> URIs. But we cannot currently negotiate the meaning of this very URI.
>
> My main concern is that reducing query answering to querying a static
> representation may be too simple an approach, same as matchmaking for
> needed products is not a simple query, but often a complex communication
> process. For example, we learn of the option space by seeing the results
> to our initial queries and then typically refine our usage of the
> vocabulary.
>
> "..Language is a living organism that adapts to the development and the
> trends of society as a whole."[1]
>
> Best
>
>
> Martin
>
>
> [1] Umberto Eco in his nice preface "The Meaning of The Meaning of
> Meaning" to Ogden/Richards „The Meaning of Meaning“
>
> Alan Ruttenberg wrote:
>>
>> On Jul 11, 2008, at 6:09 PM, Booth, David (HP Software - Boston) wrote:
>>
>>>> On Jul 10, 2008, at 7:09 PM, Martin Hepp wrote:
>>>>
>>>> Current ontology infrastructure requires that we reach
>>>> consensus first. Human communication on the contrary allows
>>>> us to postpone dispute and clarification to a later point in
>>>> time in which the disagreement becomes relevant, if it ever
>>>> gets relevant.
>>>
>>> This sounds overly pessimistic to me. Yes, some things in the
>>> semantic web *do* need to be agreed in advance, such as the general
>>> rules for determining the meaning of a statement. But individual
>>> ontologies do not -- they can be developed independently and only
>>> adopted as needed -- and there is nothing to stop an application from
>>> taking a lazy evaluation approach to semantic web data just as humans
>>> do. An application could postpone determining the meaning of a
>>> particular RDF statement (which involves determining the meaning of
>>> its constituent URIs) until it is needed
>>
>> Huh? Figuring out exactly what someone meant when they said something
>> after the fact is a huge problem. In a previous job it was routine to
>> go around to the various people who documented their experiments in
>> lab books because the lab books in isolation were too difficult to
>> understand. Understanding them after the people who wrote them left
>> the company was often impossible.
>>
>> If people can't do it, why would you expect some application would?
>>
>>> , sort of like a backward chaining reasoning style: start with the
>>> goal, and then figure out what information is needed to reach that 
>>> goal.
>>
>> The problem is that the information is encoded in the language used in
>> the statement. If you don't understand the terms you can't even get at
>> the information.
>>
>>> And if a particular statement never ends up being needed, so be it.
>>
>> Sure. But if a statement *is* needed you're out of luck.
>>
>> -Alan
>>
>>
>

-- 

-----------------------------------
martin hepp, http://www.heppnetz.de
mhepp@computer.org, skype mfhepp