W3C home > Mailing lists > Public > www-rdf-comments@w3.org > October to December 2002

Re: RDF Concepts and formal inference paths ?

From: pat hayes <phayes@ai.uwf.edu>
Date: Wed, 23 Oct 2002 13:29:30 -0500
Message-Id: <p05111b0ab9dc97eff0b7@[]>
To: seth@robustai.net
Cc: "www-rdf-comments@w3.org" <www-rdf-comments@w3.org>

>Graham Klyne wrote:
>>I agree that "no formal inference path" might include non-RDF 
>>inferences, and that one might define 'B:oneOfThem' in such a way 
>>that there is a formal inference.
>>But, in this case, I think the use of English text in an 
>>rdfs:comment to convey the intended meaning makes any formal 
>>inference path rather unlikely.
>Hmmm ... it seems to me that the formal axioms for both rdfs and 
>daml have always been expressed in English rdfs:comment(s) and 
>English descriptions in specification documents.  What's the 
>difference between transcribing those into an axiom used in a formal 
>computer inference and translating "This means the same as 
>rdfs:subClassOf" into {B:oneOfThem daml:equivalentTo 
>rdfs:subClassOf}?   There is no difference.

The issue is, WHO is doing the translation from English to a 
formalism? If the meaning is supplied in English on the web, and if 
the semweb agents are to take account of that meaning, then THEY must 
be able to read the English and provide the translation. If it is 
part of a spec, then the readers will be the human software 
developers. The difference is central.

>Behind my quibble is a very important major question.  Will the 
>culture of the semantic web embrase the idea that people can coin 
>their own terms defining them with formal languages based on 
>previously defined RDF terms?  Those new terms then become part of 
>the language of the semantic web if they gain popular usage just as 
>words become part of our natural languages culture.   The inference 
>paths on those terms *are just as formal* as the inference paths on 
>terms that are exclusively defined in the rdf, rdfs, daml, and owl 
>namespaces; the only difference is that the latter is recommended by 
>the W3C and the former is not.
>Is the W3C really in the business of  recommending how we should 
>reason?  I think not.

No, of course not. You are confusing two completely different issues, 
and the confusion is potentially dangerous. What the W3C is doing 
here is providing a framework whose primary purpose is to enable just 
the kind of social 'trade' in meaning that you want, in a very simple 
way. You seem to think that providing the underlying framework is 
tantamount to Thought Control, and nothing could be further from the 
truth. But the essential technical point that you seem to fail to 
grasp is that this framework is intended for use by *PIECES OF 
SOFTWARE*, not by human beings. Of course the software is written by, 
and acts in the name of, and to further the aims of, human beings: 
but the actual detailed work of trading meanings and drawing 
conclusions on the semantic web is intended to be done by software 
agents, not by people. Human beings bring an incredible amount of 
mental machinery to bear on the task of understanding the intended 
meanings expressed in the utterances of other human beings, and much 
of this machinery has been produced by evolution over hundreds of 
millions of years. Even then, it takes around 15 years of constant 
training (which we call childhood) to be really proficient at this 
task; and still, it may depend crucially on built-in biological 
commonalities which underlie all human languages (nobody really knows 
whether humans can learn arbitrary languages). There isn't a hope in 
hell of our being able to automate this kind of ability in software 
in any of our lifetimes, so we cannot rely on it as a basic tool for 
the semantic web.


>Seth Russell

IHMC					(850)434 8903   home
40 South Alcaniz St.			(850)202 4416   office
Pensacola              			(850)202 4440   fax
FL 32501           				(850)291 0667    cell
phayes@ai.uwf.edu	          http://www.coginst.uwf.edu/~phayes
Received on Wednesday, 23 October 2002 14:29:19 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:44:01 UTC