ppy/RulesReasoningLP-planning_chat-transcript_edited_20130725b.txt ------ Chat transcript from room: ontolog_20130725 2013-07-25 GMT-08:00 [PDT] ------ [9:21] PeterYim: Welcome to the = RulesReasoningLP Program Planning Session - Thu 2013-07-25 = Session Chair: Dr. LeoObrst (ONTOLOG; MITRE)) Topic: Brainstorm and Planning for a mini-series on "Ontology, Rules, and Logic Programming for Reasoning and Applications (RulesReasoningLP) Agenda / Briefings: * Dr. LeoObrst (ONTOLOG; MITRE) - "Session Opening" * Dr. BenjaminGrosof (Benjamin Grosof & Associates; Coherent Knowledge Systems) - "RuleLog and Reasoning Applications" * Dr. HaroldBoley (RuleML) - "RuleML Technology for Rules and Reasoning" * Dr. JohnSowa (Vivomind Intelligence) - "Relating Classical and Nonmonotonic Reasoning" * Dr. HensonGraves (Algos Associates) - "The Role of Diagrams in Logic-based modeling and ontology" * Open Discussion - ALL ** I. Open Discussion on programs and topics of interest ** II. How can we best frame the discussion and organize the effort ** III. Events & Action Plans Logistics: * Refer to details on session page at: http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2013_07_25 * (if you haven't already done so) please click on "settings" (top center) and morph from "anonymous" to your RealName * Mute control: *7 to un-mute ... *6 to mute * Can't find Skype Dial pad? ** for Windows Skype users: Can't find Skype Dial pad? ... it's under the "Call" dropdown menu as "Show Dial pad" ** for Linux Skype users: please stay with (or downgrade to) Skype version 2.x for now (as a Dial pad seems to be missing on Linux-based Skype v4.x for skype-calls.) Attendees: LeoObrst (chair), AdamWyner, AdrianPaschke, AdrianWalker, AlexShkotin, AliHashemi, BenjaminGrosof, DaliaVaranka, ElieAbiLahoud, GenZou, HaroldBoley, HensonGraves, JohnSowa, KenBaclawski, PatrickMaroney, PaulFodor, PavithraKenjige, PeterYim, RayMartin, RichardMartin, SimonSpero, SteveMandl. == Proceedings: == [9:23] PeterYim: ALL: please refresh session page (if you had loaded it a little while ago) as there is a speaking order change ... JohnSowa's brief will now go before HensonGraves' [9:29] anonymous1 morphed into AdamWyner [9:30] anonymous1 morphed into PaulFodor [9:30] anonymous2 morphed into ElieAbiLahoud [9:32] anonymous morphed into GenZou [9:32] anonymous morphed into AdrianPaschke [9:32] anonymous1 morphed into SteveMandl [9:33] AdamWyner: FYI - I have the links to the slides. Thanks for the clear organisation here. [9:33] anonymous morphed into PatrickMaroney [9:34] anonymous1 morphed into AdrianWalker [9:39] PeterYim: == LeoObrst open the session ... [9:48] PeterYim: == BenjaminGrosof's brief ... [9:50] HaroldBoley: Hilog: David Warren & Michael Kifer [9:52] HaroldBoley: A similar ('syntactic sugary') higher-order language is Relfun. [10:01] AdamWyner: Good point about LP. But, then why has that view and Prolog largely disappeared from computing departments? Has it been absorbed into other programming contexts, e.g. DBs? [10:02] anonymous morphed into AliHashemi [10:02] HaroldBoley: There's a revival at some universities such as Oxford: [10:03] HaroldBoley: Georg Gottlob. [10:03] HaroldBoley: He's part of the Datalog 2.0 workshop series. [10:05] HaroldBoley: Another modernization of LP is happening with Answer Set Programming (ASP). [10:06] HaroldBoley: E.g. Tech University of Vienna: Thomas Eiter and others. [10:08] AdrianPaschke: Modern rule engines such as Prova go beyond Prolog and extend it e.g. with typed logic, metadata for rules etc. [10:09] PaulFodor: Many universities used to have undergraduate Prolog courses, but now Prolog is just included as part of the "programming languages paradigms" course [10:13] BenjaminGrosof: LP without functions is covered all the time in DB courses; it's called Datalog, relational algebra, etc. [10:32] PaulFodor: Yes, LP is the basis for DBs. [10:06] PeterYim: == HaroldBoley's brief ... [10:31] HaroldBoley: phi(s,M) often written M |= s (for procedures: M |- s). [10:31] PeterYim: == JohnSowa's brief ... [10:32] anonymous morphed into PavithraKenjige [10:33] PavithraKenjige: But in Slide 5, every flying thing is a bird - is a wrong assumption and not correct logic! [10:36] AdamWyner: Question on Sowa's presentation. I have been under the impression that nonmonotonic reasoning cannot be reduced to KB update (adding rules, adding exceptions, constraining quantifiers) for in principle it would not be possible to produce all such updates. Sowa seems to claim that updates underly nonmonotonic reasoning. Is there a difference in view on this matter? [10:37] HaroldBoley: The egg-yolk for Chairs corresponds to what the late Wittgenstein has discussed as "family resemblances". [10:37] AlexShkotin: But there is a Frobenius theorem about number of numbers :-) [10:41] AdamWyner: Makinson (2005) is not, so far as I am aware, widely adopted by the computational models of argumentation community. Should this be promoted further? [10:45] HaroldBoley: Maybe because there are too *many* candidate bridges, so not always clear which one to use. [10:46] HaroldBoley: (Using a variety of bridges, case by case, could become messy.) [11:10] HensonGraves: Non-monotonic logic can be represented by changing the model theory to be functions defined on some base space. How does that fit with what John said? [11:13] BenjaminGrosof: Comment on JohnSowa's presentation: Classical logic is good for, and the main basis for, mathematics. Insofar as we mathematically specify and describe the formal semantics and proof theory of any knowledge representation and reasoning (KRR) system, including nonmonotonic or probabilistic, classical logic thus has a close relationship to that KRR. This relates somehow to what John was calling Makinson's notion of "pragmatics" but I'd like to understand that better. [11:17] JohnSowa: Note to Benjamin: I agree that nonmon reasoning is a requirement for any practical application that deals with anything in the physical world. [11:18] JohnSowa: But my point is that the semantics of Common Logic is perfectly adequate to serve as the foundation for a society of heterogeneous agents that use a diverse variety of classical and nonmon methods of reasoning. [11:33] AdamWyner: @John - of course. My point is only that I'd welcome a thorough discussion of the pros and cons in order to understand it better and be able to present the relevant arguments to others in the argumentation community. It is a current, ongoing, unsettled discussion in the argumentation community. [10:48] PeterYim: == HensonGraves' brief ... [10:52] AdamWyner: I've mentioned this also to Harold about graphics. Graphics are very interesting. There is research on how successful (or not) graphics are in reasoning for users; the research is largely done in the AI tutorial literature, where reasoning support is important. Also, Rationale had a visualisation for rules that was used to represent large legal cases. However, this has been abandoned due to the cost of developing the graphics in the first instance. Something for future consideration and discussion. [10:53] HaroldBoley: diagrammatic reasoning: http://www.aaai.org/Library/Symposia/Fall/fs98-04.php [10:56] HaroldBoley: Adam, but UML is largely seen as a success. [11:02] HaroldBoley: Adam, UML's Object Constraint Language (OCL) was proposed as a bridge between rules and software engineering. [11:19] RayMartin: am reading Dr. Graves slides, my apologies, did not see for whom the mini-series results are intended. [10:52] PeterYim: == moving on to the open discussion ... [10:56] PeterYim: from ToddSchneider (who sent his regrets and this suggestion): please also consider "distributed reasoning", or at least non-traditional approaches. And of course, reasoning capabilities in the context of (large) operational systems. [10:57] HaroldBoley: Rule Responder (based on Reaction RuleML) would provide an example of distributed reasoning. [11:23] AdrianPaschke: the whole topic of distributed reasoning and agents / inference services is highly relevant and has not been addressed so far [11:24] AdrianPaschke: for instance rule-based complex event processing systems are typically distriubted [11:00] HensonGraves: Is the consensus that the miniseries primary seen as a collection of tutorials? [11:04] BenjaminGrosof: Here's one idea on a first stage: 1. a session surveying use cases and applications, to motivate sessions 2-4. 2. a session on concepts and foundations of Rulelog and LP 3. a session on concepts and foundations of Common Logic, with discussion of relationship to Rulelog 4. a session on RuleML as an umbrella for standards based interoperability between Rulelog, Common Logic, and other systems such as RIF dialects, Prova, Jena, SWRL, Drools, production rule systems, Event-Condition-Action rule systems. [11:06] BenjaminGrosof: Here's an idea on a second stage/wave: cover in more depth: visualization, uses cases and applications, methods and tools for reasoning, knowledge translation, UI, and previous "non-rule" approaches to ontology modeling such as ODM and UML. [11:08] BenjaminGrosof: Also in second stage/wave: NL authoring, explanation [11:09] BenjaminGrosof: In third stage/wave: tighter and more scalable combination of logical with probabilistic/statistical ... including for data analysis/mining [11:21] BenjaminGrosof: Additional comments about mini-series planning: 1) the existing Rulelog and Semantic Web Rules conference tutorial material would correspond to approximately 3 sessions of Ontolog Forum. 2) NL-logic relationship could itself be the topic of a whole mini-series. [11:15] HaroldBoley: We could sequentialize the mini-series contributions by following the expressiveness layering of the family of rules, bottom-up (from Datalog to FOL) for Deliberation rules. Similarly, for Reaction rules (from EA and CA rules to ECAP rules with CEP), which are even more popular -- than RelDBs views (Datalog rules) -- in industry (because they deal with real-world events and actions). [11:14] LeoObrst: I would also suggest that there are some topics not directly represented here yet, and some possible speakers for future sessions: e.g., Semantic Web reasoners such as Pellet, AllegroGraph, etc., and so folks like BijanParsia, IanHorrocks, PascalHitzler, etc.; but also "distributed reasoning" as Todd suggested, e.g., Paul Tarau's work through the years on various Prologs, etc. [11:26] BenjaminGrosof: Distributed (-ness of) reasoning is indeed important, and pervasive. It includes not just interoperability (standards, knowledge translation, etc.) but also conceptually and algorithmically. [11:25] LeoObrst: What about optimization issues in ontology/rule reasoning, for efficient runtime reasoning using ontologies/rules? [11:04] LeoObrst / HaroldBoley / PeterYim: [session suggestion] do a joint session between this mini-series and the "Ontology-based Standards" mini-series (ref. last week's planning - http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2013_07_18 ) [11:04] AdrianPaschke: in terms of standards RuleML should be addressed [11:08] HensonGraves: There are a lot topics related to the talks given today. For example, An engineering model seems to me best viewed as embedded in an axiom set, rather than a Tarskian model. Engineering models often have lots of models other than what the author of the model thought he/she that they were describing or specifying. [11:11] AliHashemi: @Henson, an observation: by identifying the connection between engineering models and tarskian models, it is also possible to bootstrap ontologies in a given language. [11:10] AlexShkotin: Is it possible to take some application area (some science for ex. biology/petrology) and look at different languages to express this particular knowledge. [11:16] HensonGraves: I would like to see the mini-series take a specific example such as a chemical molecule and show how can be represented in different frameworks [11:17] AlexShkotin: @Henson, exactly! [11:17] HaroldBoley: And rules *over* a molecule/cell representation? [11:17] HensonGraves: @Alex, yes [11:25] BenjaminGrosof: @Henson: I think having a single example (e.g., molecule) as main focus is probably premature. We need to share and familiarize ourselves with identify the range of approaches and aspects and issues. Then we could hope to get to such a strong focus. But as an earlier step, having one or a few examples as a common focus for several approaches to formalize, in order to improve communication and clarity, would be very helpful and is an excellent idea. [11:19] AdamWyner: I'd be available to help with some organisation. My interests are in the Logic<->NL side and applications. The topics of the relationship of Nonmonotonic <-> Classical Logic, current logic formalisations that are feature rich, and Mark Up languages. [11:23] PeterYim: @AdamWyner - thanks for the offer to help, I'll be in touch with you offline. [11:19] AdamWyner: Seems to me that some choices will have to be made about breadth v depth of languages/varieties/intertranslation. If we could identify some core issues and samples, that would be helpful. [11:20] PeterYim: @HensonGraves, @AlexShkotin & All: HaroldBoley / PeterYim: what do you think of the earlier VinayChaudhri presentation - http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2013_06_27 [11:23] AlexShkotin: @PeterYim, It was great and it is interesting to see other languages for biology and integration tools. [11:21] AdamWyner: Problem with using science examples is that much nonmonotonic reasoning is filtered out, so is not especially relevant to reasoning in many domains, e.g. law. [11:20] BenjaminGrosof: Another important approach to relate to conceptually is Answer Set Programs (ASP). ASP is sorta in between classical-logic and Rulelog, in terms of its semantic behavior and computational complexity. [11:21] JohnSowa: ASP is an excellent approach, which can also be related to CL semantics by the same methods. [11:21] HaroldBoley: We could use subsets of HALO's Bio_KB_101 as use cases, starting from a simplified cell example, then work to more complex ones. [11:24] HaroldBoley: http://www.ai.sri.com/~halo/public/exported-kb/biokb.html [11:22] JohnSowa: I plan to update my slides with responses to the issues in the chat. [subsequent] JohnSowa:I have changed the title to "Knowledge Sharing Among Heterogeneous Agents". This new title reflects the ultimate goal. The previous title, "Relating Classical and Nonmonotonic Logic", is an important subgoal that is required to meet the primary goal. I decided that the purpose of my talk would be much clearer if I explained the larger goal before getting to the subgoal - see: http://www.jfsowa.com/talks/sharing.pdf - The major change is to add four new slides at the front to explain the ultimate goal, as expressed by the title. ... I also plan to add more slides at the end that will go into more detail about how to achieve the primary goal and the subgoal. [subsequent] JohnSowa: I received an offline question: "If all methods of reasoning about the physical world must be nonmonotonic, why should we relate them to classical logic? Can’t we just adopt one general method of nonmonotonic logic as the standard?" ... To that, my answer is: There is no ideal method of nonclassical reasoning. * There is an open-ended variety of nonmonotonic reasoning methods that have different advantages and disadvantages. * There are also many different methods of heuristic, probable, statistical, approximate, case-based, and fuzzy reasoning. * They have only one thing in common: Every nonclassical method is defined in terms of some version of classical FOL. * The only general way to relate conclusions derived by different nonclassical methods is to relate them via classical semantics. Classical FOL is an island of stability in an ocean of nonclassical methods of reasoning. [11:23] AdamWyner: @JohnSowa - my comment about Makinson was not whether his *approach* is suitably general, but the *adoption* by the formal argumentation community. [11:24] JohnSowa: For relationships between NLP, logic, and cognitive science, see my slides at http://www.jfsowa.com/talks/goal.pdf [11:25] JohnSowa: Note to Adam: That's true, but irrelevant. My recommendation was not about what has been done, but about what should be done. [11:25] AlexShkotin: May be another topic - do we have knowledge integration accumulation points for different sciences. [11:26] HaroldBoley: Alex, I think eScience is much about this. [11:27] HaroldBoley: E.g. W3C's Interest Group on Healthcare and Life Sciences - http://www.w3.org/blog/hcls/ [11:29] AlexShkotin: @Harold, is it possible to hear them here? [11:29] HaroldBoley: EricPrudhommeaux (W3C) likely yes. ... Eric also accepted an invite to an API4KB meeting. [11:32] HaroldBoley: @Alex, have a look at http://www.w3.org/2012/Talks/0717-egp-API4KB/#%281%29 ... via http://www.omgwiki.org/API4KB/doku.php?id=start [11:33] AlexShkotin: @Harold, I will! [11:29] AdrianPaschke: there is also an interesting overlap between rules and ontologies in the area of reaction rules, event/actions temporal reasoning etc. [11:29] AdrianPaschke: Reaction RuleML for instance makes use of event/action ontologies [11:31] AdrianPaschke: and non-monotonic reasoning is a requirement in reaction rules [11:32] JohnSowa: Adrian, reaction rules can also be related to the same model theoretic semantics. But it requires an additional ontology for time and causality. [11:34] JohnSowa: [referring to ???] Adrian, I completely agree. But that is still consistent with my central point: you can use a very general semantics, such as CL, for everything. [11:32] ElieAbiLahoud: I have to drop off, thx - looking forward for upcoming sessions [11:31] RayMartin: [referring to HensonGraves' response to his earlier question on who the audience of the mini-series might be] not exactly answered. sometimes one addresses the guy who wants to drive the car to the grocery store, another level to address is the guy who is building components for a vehicle, or you may be only addressing the needs of experts running an Indy car, if you will allow my simple analogy. I am not able to see who the results of the mini-series will address. [11:33] HensonGraves: Ray, to use your analogy, somewhere in between, not just users, but people who are concerned with what kind of car is needed for specific kinds of activities. [11:33] RayMartin: thanx Dr. Graves [11:33] AdrianPaschke: bye [11:33] AdamWyner: Thanks bye. [11:34] PaulFodor: Bye [11:34] PeterYim: [consensus]: adopted the title and moniker for this mini-series: "Ontology, Rules, and Logic Programming for Reasoning and Applications" and "RulesReasoningLP" [11:28] PeterYim: we will continue via email to plan out the mini-series ... I'll connect up with everyone who has shown interest in help organize [11:34] LeoObrst: Thanks all! Good session. Let's continue planning in email. [11:34] AlexShkotin: Bye [11:34] PeterYim: Very productive session ... thank you for the contribution, panelists and everyone! [11:34] PeterYim: -- session ended: 11:34am PDT -- ------