Re: Stipulative Ontologies

>A consideration, relevant to these discussions, is the degree of
>control desired by the author over the meaning of a semantic web document. 
>It seems to me desirable to add a feature, or to create an application
>that allows an author to publish how much of an ontology should be treated
>as stipulative and how much as lexical or descriptive.

I don't think that this distinction really has any meaning for formal 
ontologies. All the meaning on any formal notation is in a sense 
stipulative, surely (?) since there is no prior basis for any other 
meaning it might have. OWL is not a natural language.

>I believe the lack of this is at the heart of some of the debate on this list.
>I think that some users of RDF assumed that by default most ontologies would
>be taken as stipulative definitions of terms (URIs) owned by the author. 
>Others saw that much of natural language was based on uses that could only
>later be formalized into descriptive definitions instead.

Yes, quite: natural language. But OWL is not ... (see above).
And just for the record, nothing in any RDF, RDFS or OWL can really 
be called a 'definition'. All of these language consist of is 
assertions. Someone else can always add some more assertions.

>They were afraid
>that we were headed down a road requiring software to treat all ontologies as
>stipulative, and thus missing the chance to create terms that could evolve
>naturally.

Well, the original owner's assertions can be stipulative about what 
it is that they assert, but that doesn't stop the meanings being 
extended by others.

>
>As an example of what I mean by the use of stipulative definitions, consider
>RFC 2119, http://www.ietf.org/rfc/rfc2119.txt.  This is surely one of the most
>commonly cited RFCs.  Its sole purpose is to fix the meaning of certain terms
>when those terms are used in other documents.  It makes explicit a high degree
>of commitment to a particular interpretation of the terms used. See from that
>document the following section:
>
>   "Authors who follow these guidelines should incorporate this phrase
>    near the beginning of their document:
>
>       The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL
>       NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED",  "MAY", and
>       "OPTIONAL" in this document are to be interpreted as described in
>       RFC 2119."
>

Right. The point of this kind of talk , and its utility (and 
similarly in legal contexts, contracts and so on) is to *limit* any 
other interpretations or ancillary meanings that these words might be 
taken to have by virtue of some aspect of their NL meaning, to guard 
against possible misinterpretation. But OWL URIrefs have no other 
potential source of meaning to guard against or to be protected from. 
Suppose one were to try to imitate this kind of language when talking 
about a URI in some OWL; what could you say? It would come out like 
this: The URIref "http://www.ihmc.us/users/phayes" in this ontology 
is to be interpreted according to the way it is used in this 
ontology. Well, yes.... but is this worth saying?

>Stipulative definitions are used quite often in statutory law, contracts,
>programming languages, and standards documents.  They are used, it seems,
>wherever the advantages of reduction of ambiguity and increased
>precision are desired.  They have disadvantages as well, and certainly
>need not be used everywhere.
>
>I believe we can formalize and thus automate the specification of the degree
>of stipulation we desire over our ontologies

I would be interested in knowing how you propose this might be done. 
I have absolutely no idea even how to begin to tackle such a project. 
Remember that OWL identifiers are URI references, not English words.

>and thus meet the needs of both
>those who think that the author/owner of URIs should be able to stipulate
>the meaning intended by publication of those URIs in certain contexts and
>those who think the meaning of URIs should be free in all contexts to evolve
>naturally towards whatever future they may have.

To publish some OWL is generally understood to be an assertion that 
the OWL is true. This might be called a partial stipulation. It is 
stipulative to the extent that it says something positive about the 
intended meaning. It is partial since it generally does not 
completely determine the intended meaning. Nothing in natural 
language is exactly like this.

Pat Hayes

>
>Having the explicit facility to do this will not inhibit either goal.
>
>John Black


-- 
---------------------------------------------------------------------
IHMC	(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32501			(850)291 0667    cell
phayes@ihmc.us       http://www.ihmc.us/users/phayes

Received on Tuesday, 6 April 2004 00:44:12 UTC