RE: Comments to Rules Working Group Charter Draft $Revision: 1.60$ (Part II)

> -----Original Message-----
> From: Dave Reynolds [mailto:der@hplb.hpl.hp.com]
> Sent: Friday, September 09, 2005 7:21 PM
> To: Giorgos Stamou
> Cc: 'Sandro Hawke'; public-rule-workshop-discuss@w3.org
> Subject: Re: Comments to Rules Working Group Charter Draft $Revision:
> 1.60$ (Part II)
> 
> Giorgos Stamou wrote:
> 
> > 2.x Uncertainty and fuzziness
> >
> >
> >
> > It would be useful for the language to be able to represent uncertain
> > and vague information. Thus, the extension of the core language with
> > uncertainty reasoning and fuzzy logic capabilities will be provided. A
> > requirement for this extension is that it should generalize the
> > two-valued Boolean logic of {0,1} into the interval [0,1], by providing
> > a sound extension of Boolean logic. Hence, such a feature should not
> > affect applications that do not require the specification of
> uncertainty.
> 
> Whilst the presence of such a feature need not affect applications that
> don't use it, developing it could be a noticeable extra burden on the
> working group. It might be better left to a later phase.

[Giorgos Stamou] 
As I explained in my previous e-mail, it is much better to seriously
consider starting this work from the beginning, since it is very important
for this effort to follow all the phases of the work done within the WG. And
this possible noticeable extra burden could be obviously supported by people
active in this area. Let me mention that the fuzzy RuleML Technical Group
has 10 active members. They are interested in the W3C new Rule Language WG,
some of them participated in the Rule Workshop in Washington and now they
are participating in the discussion about the charter. I am quite sure, that
most of them also really like to contribute to the work on the proposed
uncertainty and fuzziness language feature.

> 
> First, any solution doesn't just have to work with rules. The charter
> requires interoperability with RDF and OWL and so the WG would need to
> explain how the uncertainty representation affected those as well. Has
> there been enough work done on combining DLs with uncertainty calculi to
> be confident of the right way to do that?

[Giorgos Stamou] 
I agree that the compatibility issue with RDF and OWL is very important and
I am looking forward to see how the new charter defines this compatibility.
In any case, a lot of work has been done in the area of fuzzy description
logics and could be used in the proposed effort. Some papers are cited in my
first e-mail from which you can find several references.   

> 
> Second, whilst I agree that fuzzy logic is well understood and is suited
> to
> use in rules, is it really so clear cut that it is the correct uncertainty
> representation to standardize? The use cases described at the workshop
> seemed be evidence combination problems that I would naively have expected
> to be as well suited to, say, a probabilistic approach.

[Giorgos Stamou] 
The document I sent to Sandro leaves this issue quite open. The only
restriction is that the extension should soundly generalize the two-valued
Boolean logic, which is very important. Nevertheless, I believe that fuzzy
logic is more general and more suitable than other relevant theories to the
specific industry requirements as clearly mentioned in the Worhshop, but
this is just my opinion.
 
> 
> Perhaps you just need the charter to say that the WG should not preclude
> experimental extensions which provide an uncertainty calculus and which
> may
> be the subject of a future round of standardization.

[Giorgos Stamou] 
We just need the industry requirements expressed in the Workshop and
strongly motivated the need for uncertainty and fuzziness, to be reflected
in the charter and hopefully this issue to be put with some few words in the
scope of the WG charter as a desired language extension. 

> 
> Dave
> "... keep it simple ..."
> 
> 

Received on Friday, 9 September 2005 20:42:10 UTC