From: Benjamin Grosof (bgrosof@MIT.EDU)
Date: 04/24/03
Hi folks, Here are my notes from the discussion on our last telecon. (Mike, you might want to integrate some of this into your minutes.) Benjamin % notes from Joint Committee discussion of abstract syntax for RuleML 4/22/03 % by Benjamin Grosof % note the sequence below is reorganized (for clarity) from that of the % actual telecon discussion % the discussion revolved mostly around the more controversial aspect % of Benjamin's presentation: should we try to provide an OWL or RDF syntax % for RuleML Benjamin: an objective favoring why try for an OWL syntax: represent the ontology of RuleML's sublanguages hierarchy, then use the same toolset for processing an instance rulebase Stefan: another objective is to facilitate meta-reasoning about the rules themselves, e.g., to query for all rules whose head mentions a particular predicate P. Benjamin: another goal for why encode rules in RDF or OWL: conjecture that it will then be easier to extend RuleML to have more generic RDF or OWL for the arguments/terms within the rules Benjamin: in attempting an OWL syntax for RuleML, there are challenges of (1) closing-off, and (2) representing ordered collections in such a way that one can restrict the types of the members of the collection Benjamin and Stefan: one reason why aim for OWL rather than , but still challenging in OWL currently, is the following objective: to enable XML-Schema's kind of validation, esp.: - check that a rulebase is well-formed syntactically - detect that a rulebase falls into a particular syntactic subset wrt which features are actually used Ora: can view there being "conventions of usage" that go beyond what can be expressed in the OWL itself, e.g., wrt closing-off or orderedness Mike Dean and Peter and Pat: wrt challenge of how to represent ordered typed collections, e.g. lists of logical terms: can subclass off of rdf:list but then can't use parseTypeCollection to generate first, rest, end (? hope I understood this right) Ora: wild idea: represent checking of that in RuleML itself one possible approach: define new RDF or OWL builtins for generic containers of typed members, then define subclasses and subproperties for specific usages, without changing fundamental DL expressiveness; e.g., indexed accessors for them, and a size/length/count property; e.g., a syntactic shorthand of _1, _2, ...; subproperty of ordered container index Peter: also it's nice to detect "referential transparency", i.e., that every predicate mentioned actually has a (non-empty) definition Mike D. and Said and Benjamin: *in the rules requirements document, we should discuss the requirement that one can represent the ordering of rules within a rulebase, or the ordering of literals within a rule, since this is crucial for practical efficiency in many actual rule systems / engines - many others chimed in: yes, that seems reasonable / important ________________________________________________________________________________________________ Prof. Benjamin Grosof 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 : 04/24/03 EST