% notes of verbal remarks and discussion, and related highlights, % from Ontolog Forum session 10/24/2013 kickoff of RulesReasoningLP mini-series % by Benjamin Grosof, co-chair of session Disclaimer: These notes are NOT official nor complete nor guaranteed-accurate. I took them for personal purposes, but think many others may find them useful. %%%% Benjamin Grosof remark during intro of miniseries: - lot of focus in miniseries on how ontologies and rules come together, how rules get used for ontologies %%%% opening remarks by various (2-min. each) Michael Gruninger - advocate of expressive, esp. FOL and CL - need good support for automated reasoning to be really useful - looking for guidance from reasoning community wrt tools for ontologies to be evaluated and applied - looking for test cases from the ontology community - hope for insights on what is the appropriate ontology language to use operationally for various applications, not just original development of ontologies Leora Morgenstern - at Leidos, co. new name of split SAIC - working on DARPA seedling on how to go from regulatory text to executable rules . looking at financial, incl. SEC, and other Patriot Act - extract intermediate representations, incl. . as rich an ontology as one can get . dependency graph among rules . semantic parsing and assignment for roles incl. permission/obligation - an immediate issue is what representation to use . RIF is clearly inadequate expressively . exceptions are such a fundamental concept, incl. priorities o need language like SILK -- ie Rulelog Michael Kifer - think the rules community has the best technology for KR, but somehow not getting its due in the community - RIF - systemic problems due to the charter could not produce the dialect most useful for the rules community, had to focus on BLD which is basically Datalog and not so useful - later tried to produce more expressive dialects, which use Well Founded Semantics, but it's hard to get attention for this - now I and Benjamin are pushing Rulelog - would like people to pay more attention to standards, and produce right standards - not clear that W3C is the right venue for this, since not clear they are interested in continuing the RIF effort - feel that the efforts overall are fractured, incl. there is an effort for Answer Set Programs but they don't seem interested in non-ASP, only ASP . need to overcome this fragmentation Vinay Chaudhri - (see his slide) - separation between ontologies and rules is artificial - focus should be on decidable reasoning . folks in rule languages community have not focused so much on this, but rather on expressiveness . in DB systems, the performance guarantees are crucial Harold Boley - (see his slides) Henson Graves - (see his slide) Ken Baclawski - (see his slide) John Sowa - (see his slides, he mainly covered slide 2 verbally) - recorded (started about 10:14a PDT) - vagueness and uncertainty are important . contrast that with very clean model-theoretic kinds of approaches Mike Dean - delighted to see this cooperation b/ ontology and rules community, which have largely been disjoint despite many common goals Peter Yim - (see his slide) - feel this miniseries is a very important next step %%%% survey by Leo (see his slide deck) (Benjamin's note to self: nice/cute ex. of semantics of material implication and contraposition: "if pigs can fly ..." on slide 7) %%%% survey by Benjamin (see his slide deck) Cyc is much closer to Rulelog (vs. to FOL or to LP) %%%% general Q&A and discussion at end Hassan: presentations very good, but bit biased in that ignored some important relevant work - in particular: lambda calculus, where a function is described as a rule - that was the basis for the first work on higher-order programming . ___ in '76 PhD work . Dale Miller on lambda-Prolog . Teyjus.cs.umn.edu system work at U Minn by ___ -- see that for references - key to cover in miniseries: operational, implementation, pragmatically how run, how efficient . Prolog via WAM . lambda-calculus also implemented via a sub-mechanism - work on constraint logic programming . gives Prolog technology much bigger power, opens the LP paradigm to anything that one can represent as a constraint solving process, incl. probabilistic or fuzzy or whatever - there's a new generation that's not very educated in logic programming in general . we need to make it learnable without undue effort Benjamin in response to Hassan: agree - also very important is modern LP implementationally are tabling techniques, a kind of caching of work on subgoals and their results Peter: wrt Leo's presentation, slide 26 (missed this bit of the discussion) JoelBender: Q about description of RuleML <--> RDF ontologies as in foaf or whatever A by Harold and Benjamin: - there are relationships in terms of data models, . Harold will post some links to chat; also it's important - it's also important to understand at logical/semantic level, incl. what's essential in syntax AlexMirzaoff: Q about how things change in time - A by Leo: can use ontology of time or temporal logic - A by Benjamin: temporal reasoning is representable within expressive general logics such as CL and Rulelog; . practically, defeasibility is often very important to do temporal reasoning efficiently and tersely, particularly to represent causality, incl. projection forward in time or backward in time (as in abduction and inductive learning) %%%%