From: Benjamin Grosof (bgrosof@MIT.EDU)
Date: 06/03/03
(inline'd and attached too for convenience) % notes from JC telecon 6/3/03 % by Benjamin Grosof Benj reviewed the basic definition of Horn LP can intro extensions which are reducible to the basic: - OR in body - existential in body - AND in head - universal in head Pat: motivations for the definition and semantics? Gerd: natural for executable to have a single intended [partial] model, similar to databases Pat: Herbrand is useful for meta-theoretic and syntactic Benj: has simplicity for implementation and also for specification Pat: these questions will reemerge Mike: e.g., how should it relate to semantics of RDF or OWL, as a target - Pat: RDF is basically existential conjunctive logic - Benj: LP is actually much more established than DL or RDF or OWL - Mike: try to conform to OWL, head off any incompatibilities Pat: differences between LP and OWL/RDF include: RDF and OWL all have classical model theories LP is fundamentally not monotonic and is recursive; it plays a different game Gerd: would not be as pessimistic as you Benj: don't get hung up on the terminology of "model": the Horn LP conclusion set corresponds to the set of ground atoms that are entailed in Horn FOL Pat: example: RDF or OWL ontology that talks about colors, e.g., red, green, blue, that is integrated with PanTone (spelling?) which has URI's, and printing and a color-checking engine; in Horn LP model theory a name refers to itself, so can't express this well; will wreck things. Benj: elaborate please? Pat: point is that URI's have genuine denotations in the world Benj: that's orthogonal to Herbrand-ness unless you can represent within the KR some info relating the denoter to the denotation Ian: why stop at Horn LP rather than Horn FOL? Benj: because it's well understood at a research level how to extend it to non-monotonicity and procedural attachments, but not to do so by extending FOL (tho' some procedural attachments can be handled in FOL). Also Horn LP is a waystation not a final destination -- it gets extended in two directions: towards FOL/classical-logic and also towards fuller LP with nonmon and procedural attachments which is commonly used practically and relatively well understood theoretically. Gerd: there is a generally used notion of rule systems, that is non-classical, e.g., has negation-as-failure Pat: I'd like to get convinced that logic programs is the right notion that is widely used [in commercial rule systems] -- feel doubtful on this point general discussion on this point - Benj, Gerd, Said observing that Said: policies Ian: lots of these systems mix in programming application development features, don't they Said and Benj: yes Benj: but they're closer to declarative LP KR when you project onto their pure knowledge-based aspect consensus: important topics for future discussion: - built-ins (next week) - human-consumption string syntaxes (2 weeks from now) - explicit equality, names (3 weeks from now) - semantics of procedural attachments, how to relate that to classical logic ________________________________________________________________________________________________ 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 : 06/03/03 EST