From: Benjamin Grosof (bgrosof@MIT.EDU)
Date: 06/08/04
________________________________________________________________________________________________ Prof. Benjamin Grosof % notes from Joint Committee telecon 6/8/04 % notes by Benjamin Grosof participants: - Mike Dean - Deborah - Norman - Harold - Said - Benjamin - Sandro - Pat o news of SWSL [Benj] - upcoming F2F there for about July 16, plan to use RuleML nonmon LP and FOL - would be nice by then for us to have spec on FOL and of updated RuleML o news of WWW Conf. [Mike, Benj, Harold] - DevDay Rules on Web track was a highlight - relevant themes: use of rules for services, combo of rules with ontologies o main agenda: discussion of FOL strawmen from Benj and Said and from Pat Benj and Said presented theirs Benj: started top down, half-way worked out now aimed to support XML serialization done in style similar to previous RuleML abstract syntax Benj A to Mike/Pat Q: intended to be just straightforward classical FOL Benj: issue: how to treat closed-ness? - tried to do that; Benj: think strawman needs fixing actually in this regard Pat: top-level in SCL every name not quantified over is treated as a constant issue: do we want free variables? - query lang could be something separate, and query variables could be viewed as existentially quantified at outer scope Benj: issue: how do we represent in the grammar the sufficiency of binding? I.e., so that XML parser etc. tools can do it? Seems we have to specify it as a separate constraint for implementers - general discussion about this, seemed to agree there's no other way. Pat: if you can recognize variables as a separate syntactic category Benj: I propose: - have syntactically explicit variables, as in current RuleML and SWRL - have perhaps an ability to declare in a collection of FOL statements a convention for implicit universal (or perhaps also optionally existential) quantification at outermost scope -- or perhaps also optionally no quantification, i.e., treatment as free variables Q by Mike: relationship to RuleML? A by Harold: we could introduce explicit quantifiers and also disjunction into RuleML. A by Benj: yes, and more cf. Lloyd-Topor. For both the reducible-to-LP-expressiveness base (with mon, nonmon subcases) and the general-FOL cases. Actually, that's been our plan in RuleML all along. Pat: want to label an expression as fitting into / naming various expressive/syntactic classes, i.e., tag expressions with those names Benj: yes. We should have an ontology of expressive/syntactic attributes and combinations, and be able to annotate/label with that at the grain of whole KB/rulebase or at the grain of individual rule/axiom/clause/expression. Pat and Benj: The tagging should be open ended, including so that can help implementers declare constraints that facilitate processing. - Pat: Let others define the relevant expressive/syntactic subsets. Pat: key people working on SCL spec actively now are myself, Tanel Tammet, and Chris Menzel. Pat: John Sowa now has a nice controlled English that maps to FOL. Harold: issue: making DL folks happy - Pat: we can view DL as subset of FOL, so that hopefully will Pat: a recent proposal in SCL is that one can specify a special constraint that keeps it to classical FOL, i.e., mark all predicates or functions as not part of the domain of quantification - Benj: that's rather blow by blow, it would be nice also to specify that at the grain of the whole KB - Pat: yes Harold: what's the status of XML syntax for SCL? Pat: there's been a lot of arguments about that. Murray Altheim (at Open Univ. in England, background in Topic Maps) won the argument since he knows the most about it. The XML will be designed to be extensible rather than beautiful. Benj and Harold: let's have RuleML and SCL efforts join forces sooner rather than later - Pat: yes, I'll plan to start with a strawman Benj Q: in SCL: what's rationale for equation being top level? To treat Pat A: there was a lot of philosophical baggage. A key issue was that equality symbol should not be treated as part of domain of quantification. So essentially equality is treated specially syntactically -- it's essentially a first order language with equality. It has a fixed meaning -- a=b means that a and b denote the same thing. Intended as common garden equality. It applies to relations too; there, sameness of extension doesn't imply (necessarily) equality. Benj: wrt plan -- how to go forward? Would be nice to have a single strawman, using the above ideas. Pat: would be nice to get comments on SCL Benj: let's try to converge efforts, i.e., RuleML/SWRL inspired XML syntax for SCL. It could turn into an alternative or even main XML syntax for SCL. Pat: sounds great! - there's a year-old design document on guidelines and concerns on XML syntax, from Murray/SCL folks, I'll send that out to the committee list Pat: warning: SCL spirit tends to shrug shoulders about ability to check on things like arity etc. - Benj: we could start out with a somewhat more constrained version that facilitates implementation, e.g., for services - Pat: yes Web Technologies for E-Commerce, Business Policies, E-Contracting, Rules, XML, Agents, Semantic Web Services MIT Sloan School of Management, Information Technology group http://ebusiness.mit.edu/bgrosof or http://www.mit.edu/~bgrosof
This archive was generated by hypermail 2.1.4 : 06/08/04 EST