Re: NAF v. SNAF - where is this being addressed?

Jim Hendler wrote:

>  Seems to me if I see your rule set includes a NAF-based rule, and you 
> give me a conclusion to something, that if I don't know what graph/KB/DB 
> that was applied to, then I have no way to know whether I can use your 
> result in my application

Absolutely.  But in general, I can't take a rule set, with or without 
NAF, and use it in my application without some knowledge of the 
assumptions underlying it.  As Pat Hayes says, your universe may be 
different from my universe, and so the same argument applies to 
universal quantifiers.

The point of SNAF is that if I know what reference KB we are assuming 
for the ruleset, I can agree that the use of NAF makes sense.  And there 
is a corresponding "SUQ" concept for the interpretation of "for all x".

>   Seems to me also that this has a big effect on the charter, as I don't 
> know if there is an agreed upon use of SNAF for the Web, and would need 
> to be something the WG would be required to elucidate.

Well, I agree that the scope of any rule using NAF needs to be identified.

But it is clear that there are "KBs" that are taken, across the entire 
Web, to be the reference repositories for certain "fact types", and the 
use of NAF on such a repository can be treated as universally 
appropriate.  For example, no one should doubt the validity of NAF 
reasoning over the IANA list of Internet addresses, or the W3C list of 
W3C Recommendations.

> p.s. Note that in datalog, there is always the assumption that the rules 
> and a particular database can be linked - on the Web, that is not 
> necessarily true.

I would argue that this is not quite the right consideration.  A common 
set of "business rules", for example, may be applied over different 
data/knowledge bases in different business units of the organization 
responsible for the rules.  Similarly, a common set of accounting rules 
may be legally required for all businesses in a particular governmental 
classification, even though every such business has its own database. 
And those rules may include NAF rules, e.g. "if no payment has been 
recorded after 30 days, the account is 'delinquent'".

Datalog rulesets are clearly associated to a particular database for 
"execution".  They are usually tightly coupled to a particular DB 
schema, which greatly reduces the possibility of wider applicability. 
But Web rulesets will also be coupled to a particular database for 
"execution", just as Web ontologies will be coupled to a particular 
knowledge base for "reasoning about individuals".  The difference is 
that Web rulesets are based on some reference ontology, and both the 
ontology and the ruleset must be "interpreted" into the vernacular of 
the target knowledge base in order to be "executed".  So the question is 
whether the ontology and ruleset are *appropriate* to the specific 
knowledge base over which you want to reason.  Whether that is the same 
knowledge base I am using, or my trading partner is using, may or nay 
not be important.  But if it is important that they be the same, i.e. 
that there is a reference knowledge base, either for all such reasoning, 
or just for our joint business/research, that is a separate issue.

-Ed

-- 
Edward J. Barkmeyer                        Email: edbark@nist.gov
National Institute of Standards & Technology
Manufacturing Systems Integration Division
100 Bureau Drive, Stop 8264                Tel: +1 301-975-3528
Gaithersburg, MD 20899-8264                FAX: +1 301-975-4694

"The opinions expressed above do not reflect consensus of NIST,
  and have not been reviewed by any Government authority."

Received on Wednesday, 29 June 2005 16:35:50 UTC