Re: Some Comments on RIF BLD from an external reviewer

Yuh-Jong,

Thank you for taking the time to provide us with feedback.

Yuh-Jong Hu wrote:
 > As a privacy protection and digital rights management policy researcher, I
 > am happy to see RIF working group delivers the RIF BLD (Basic Logic Dialect)
 > working draft (http://www.w3.org/TR/rif-bld/ ) and other related working
 > drafts, such as RIF Framework for Logic Dialects (FLD) and RIF RDF and OWL
 > Compatibility. This indicates an important step for standardizing the rule
 > exchange language for semantic web technologies and related applications.
 > After quick examing the RIF BLD working draft, I would like to provide some
 > comments that need to be clarified:
 >
 >
 >    1. Do we need to webize the  terms from rule module as well as ontology
 >    module in the RIF BLD? This webizing issue for both condition language and
 >    rule language can be further enhanced in the near future. For example, in
 >    EBNF grammar, it does have IRI (extension of URI) to refer to the original
 >    sources of FORMULA and ATOMIC terms. The naming prefix in the document/group
 >    preamble also provides the rule modules import sources. But it is quite
 >    possible some of imported terms in both the condition language and the rule
 >    language might come from ontology modules specified as RDF(S) and OWL
 >    ontology langauge. In that case, these terms and expressions might need to
 >    be webized. I am wondering whehter RIF BLD spec. itself already has this
 >    capacity to declare and webize the ontology related terms? Maybe, we can
 >    refer to the RIF RDF and OWL compatibility working draft to resolve this
 >    issue (I check this draft and it only very briefly discuss in section 5). On
 >    the other hand, we might said this is a rule interchange specification so we
 >    do not have to consider webizing the terms from ontology. But both condition
 >    langauge and rule language do have member and subclass terms in their
 >    respective EBNF grammar, I am not quite sure how these terms will be applied
 >    in the rule language expression because the membership and subclass term
 >    relationships usually happen in the ontology schema. Therefore, this also
 >    reflects the necessarity of webizing ontology terms in the condition/rule
 >    language.

I'm not sure what you mean by "web-izing". Do you mean assigning a URI to 
constants (e.g. individuals, predicate & function names)? If so, then yes, BLD 
allows you to use URIs as terms.

However, if these URI terms are also defined in an ontology (RDFS or OWL), then 
you have to specify how to combine the two kinds of semantics. The details and 
parameters of such combinations are specified in the RDF&OWL Compatibility document.

 >    2. Does the RIF-BLD provides the power to resolve the semantics
 >    discrepancy from underlying different logic program-based rule systems? As
 >    mentioned in RIF BLD working draft 1 overview, for those people who need a
 >    direct path to implement but are not interested in extensibility issues,
 >    they choose RIF-BLD. Otherwise, they choose  a specifialization of the
 >    RIF-FLD. In this case, I am not sure whether RIF-BLD might be quite limited
 >    when we need to consider a variety of rule systems which might have the
 >    expressive power beyond RIF-BLD. The decision for people to choose RIF-BLD
 >    or RIF-FLD might need to be more clear and specific in the working draft.
 >    For example, if one rule interchange system is based on RIF-BLD and another
 >    rule interchange system is based on a specialization of extensible RIF-FLD,
 >    can those respecitve underlying rule systems with RIF-BLD and RIF-FLD as the
 >    RIF language in their hub centers and interchange their rule modules without
 >    worrying any semantic ambiquity problem after integration?

You have the right idea. BLD is intended to be able to express the kinds of 
rules that are expressible in most (logic-based) rule languages. Thus BLD does 
not have the power to resolve semantic discrepencies between different LP 
systems, quite the opposite - it avoids that by providing syntax&semantics only 
for the kinds of rules for which there are no discrepencies in different systems.

So if you have a rule language more expressive than BLD (as most rule languages 
are), then even if it is specified as a specialization of FLD, only certain 
rules will be interchangeable through BLD.

It's important to realize that BLD does not solve the general problem of rule 
interchange, it takes a step in that direction.

 > Overall speaking, all of RIF-BLD and related working drafts do consider the
 > semantic web cutting edge technology and development status and I am
 > expecting these RIF standards can be finialized in a near future.

Thank you for the comments.

-The RIF WG

-- 
Dr. Christopher A. Welty                    IBM Watson Research Center
+1.914.784.7055                             19 Skyline Dr.
cawelty@gmail.com                           Hawthorne, NY 10532
http://www.research.ibm.com/people/w/welty

Received on Monday, 29 September 2008 02:08:56 UTC