ppy/chat-transcript_unedited_20120301a.txt (rev.1) ---------- Chat transcript from room: summit_20120301 2012-03-01 GMT-08:00 ------------ [09:11] PeterYim: Welcome to the = OntologySummit2012: Session-08 - Thursday 2012-03-01 = Summit Theme: OntologySummit2012: "Ontology for Big Systems" Track (Cross-Track-A2) & Session Title: Ontology for Federation and Integration of Systems Session Co-chair: Mr. CoryCasanave (Model Driven Solutions) and Mr. AnatolyLevenchuk (TechInvestLab) Panelists ... with briefings from: * Mr. DennisWisnosky et al. (DoD, US) - Leveraging Semantic Technology across systems to meet the goal of having an "executable, integrated, consumable, solution architecture" * Dr. TommiKarhela (VTT Technical Research Centre, FI) - "Ontology based Integration Platform for Modelling and Simulation - Simantics" * Mr. AnatolyLevenchuk (TechInvestLab, RU) - "Ontology-based Systems Federation" * Mr. CoryCasanave (Model Driven Solutions, US) - "Semantic Information Modeling for Federation" Session page: http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2012_03_01 Mute control: *7 to un-mute ... *6 to mute (please make sure your own phone is not muted as well) Can't find Skype Dial pad? ... it's under the "Call" dropdown menu as "Show Dial pad" . == Proceedings: == . [09:19] anonymous morphed into Damian Gessler [09:19] Damian Gessler morphed into DamianGessler [09:23] anonymous morphed into Tommi Karhela [09:26] anonymous morphed into GeorgeDobbs [09:26] anonymous1 morphed into Sjir Nijssen [09:27] anonymous morphed into Sjir Nijssen1 [09:28] anonymous2 morphed into JohnBilmanis [09:28] anonymous1 morphed into MatthewHettinger [09:31] anonymous1 morphed into KenAllgood [09:31] anonymous morphed into AndreyBayda [09:31] anonymous1 morphed into ChristopherSpottiswoode [09:32] anonymous morphed into CDR Chris Harris, JNWC [09:33] anonymous morphed into AlanWu [09:33] anonymous1 morphed into Wen Zhu [09:34] Harold Boley: Hi. [09:34] anonymous morphed into DavidEddy [09:35] anonymous morphed into TomTinsley [09:36] Room password was updated by: Sjir Nijssen [09:39] AliHashemi1 morphed into AliHashemi [09:39] JoelBender: (test message) [09:41] PeterYim1: for some unknown reason ... someone caused a password to be put into the session (and is messing everything up) [09:41] anonymous morphed into ElizabethFlorescu [09:41] PeterYim1: the password is "ontolog" ... I have added that to the session page ... hope people can see it ... I will try to get a chance to announce that on the call [09:42] AliHashemi1: @Peter - Maybe we should make a voice announcement in the call [09:42] AliHashemi1: I had to continually refresh the session page and just barely noticed the password [09:42] AliHashemi1: (perhaps bold it on the session page as well) [09:43] AliHashemi1 morphed into AliHashemi [09:44] Victor Agroskin1 morphed into Victor Agroskin [09:44] anonymous morphed into NicolaGuarino [09:47] anonymous morphed into DeborahMacPherson [09:47] Victor Agroskin morphed into VictorAgroskin [09:48] PaulWitherell1 morphed into PaulWitherell [09:49] anonymous morphed into BobSchloss [09:49] SimonSpero1 morphed into SimonSpero [09:50] PeterYim1: @Cory - can you please find a chance to tell people about this mishap on the chat-workspace, and that they should get back in with the password "ontolog" [09:50] AmandaVizedom1 morphed into AmandaVizedom [09:51] anonymous morphed into AvronBarr [09:51] CoryCasanave: @Peter, will do - suggest you also put it on the screen. [09:52] anonymous morphed into Ed Seidewitz [09:52] AnatolyLevenchuk1 morphed into AnatolyLevenchuk [09:53] PeterYim1: @Cory - I have noted it on the session page, but people need to refresh the page to see it - http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2012_03_01#nid35AU - AliHashemi has even helped to highlight it too (thanks, Ali) [09:54] PeterYim1: @ALL - slides for the DennisWisnosky talk has now been posted - please refresh you session page to see it [09:55] PeterYim1 morphed into PeterYim [09:55] AlanWu: test. please ignore [09:56] AmandaVizedom: Thank you, Cory & Peter [09:56] FabianNeuhaus1 morphed into FabianNeuhaus [09:57] PeterYim: == TommiKarhela presenting ... [09:57] anonymous morphed into KathyEllis [09:58] PeterYim: TommiKarhela: there should be a more up-to-date slides will be made available (and will be available on the session page) [09:59] PeterYim: == AnatolyLevenchuk presenting ... [09:59] NicolaGuarino: @Tommi: the simulation models you mentioned are mathematical models, are they? [10:01] DeborahMacPherson: Need to sign off for an hour for a conference call, will be back [10:02] Martin Serrano - DERI1 morphed into Martin Serrano - DERI [10:04] NicolaGuarino: @Anatoly: still I miss the difference (if any) between federation, integration, sharing, interoperability.... [10:05] Martin Serrano - DERI1 morphed into Martin Serrano - DERI [10:05] MatthewWest: @Nicola: There isn,t really any. Just words preferred by different communities. [10:06] CoryCasanave: @Nicola - Independent control, conception and life-cyle are markers of federation where as integration can involve more contol over the parts to be integrated. [10:07] NicolaGuarino: @Cory: I agree. Indeed in my understanding federation is weaker, in a sense, than integration [10:07] CoryCasanave: @Nicola, I would say less coupled [10:07] NicolaGuarino: Yes [10:09] JamesOdell: So, is Anatoly saying that the "corporate cyborg" is agent based? [10:12] BobSchloss: I think Anatoly is saying that if you wanted to determine the extent of interoperability between two software stacks, perhaps operated by two different enterprises, you need to think about structure, process/workflow, APIs, visible data (for fetch or deposit). At least this is something I have been thinking about. Further, I think Anatogy is saying that different kinds of models have typically been built by different stakeholders in the enterprise, and the alignment between these different types of models hasn,t been address in a systematic way (in most organizations) in the last decades... but he thinks it is perhaps now becoming possible. [10:13] CoryCasanave: @Jim - is an agent and "agent based" are the mae thing? [10:13] JamesOdell: @Cory - yes [10:16] BobSchloss: Observation: Most of the "web-scale" evolvable software systems have used principles of "loose coupling" and "lowest permissable cross-understanding of the semantics of data entities". Anatoly appears to want to not have to rely on those pragmatic approaches... but to develop a new "web-scale" approach to ontologies-and-modeling-and-semantic-oriented-programming. [10:16] JamesOdell: Since processes can be defined as concepts in an ontology, a formal process model can in the ontology, therefore, makes the ontology "executable." [10:17] BobbinTeegarden: @Jim ,executable, as in a process? or ,executable, in that an agent can execute a process (agent as uri)? [10:19] BobbinTeegarden: @Tommi Layer0 is the ,language, you use, is that a programming level, API, and does it back into RDF/OWL? [10:21] JamesOdell: @Bobbin - With ontology, you can either/both techniques. (IMO, agent based processing would be more manageable and maintainable.) [10:24] NicolaGuarino: @Anatoly: what is the ISO 24... enterprise modelling standard you mentioned? [10:24] anonymous1 morphed into ChristopherSpottiswoode [10:27] FabianNeuhaus1 morphed into FabianNeuhaus [10:28] BobbinTeegarden: @Anatoly seems like upper (vs lower) ontology is two layered/binary, while it seems you are hinting that ontologies are more fractal in shapes -- layers decomposed or abstracted -- almost a continuum of abstractions; and that you need to work with multiple levels of abstraction at once? Is that where you were going? [10:28] BobSchloss: What Anatlogy says disturbs me; it ignores much of what the commercial software world has learned about "agile software engineering" and is now trying to extend into "agile (incremental) modeling". I suppose certain classes of customers (such as people who design and operate nuclear power plants) can really do this complex multi-aspect modeling, but that seems like a relatively small part of the engineering-dependent business sector of the economy. [10:31] PeterYim: @ALL: updated slides from TommiKarhela has already been posted [10:32] CoryCasanave: @Bob, what would be an example of a solution that is solved with only one viewpoint? [10:34] AmandaVizedom: I think that it is possible to teach some people the foundations of ontology engineering quite quickly. One reason short training often fails is that everyone (students, instructors, stakeholders paying for the training) wants to skip straight to the language and tools to be used, without going through the foundations first. That doesn,t work. Without learning the foundations, most students will not understand (and will misunderstand) the methodological content and distinctions, and will not be able to apply them successfully. While it,s true that not everyone can learn the fundamentals, many bright technical folks can if fundamentals are taught clearly and up front. And with the fundamentals, it is possible to do much more with just a little training on a specific language, tool, or project approach. [10:34] MikeBennett: This is interesting. We are taking a similar approach I think, which is to identify those industry and academically referenced ontologies which contain abstractions we need to refer to, and consider how to socialize these under a common set of under-specified "philosophical" partitions, taking care of viewpoints etc. [10:35] BobSchloss: @Cory - Few systems are solved with only one viewpoint. But the point is the modeling of aspects of the intended system is performed in an incremental way... a period of time where the focus is on processes and activities, then a period of time where candidate service interfaces are identified and a first decomposition of the system, then a period of time about externally visible information, and although you can continue to cycle through these different aspects, you have a "coherent model" after the first iteration. [10:36] CoryCasanave: @bob, would this not be a methodology for what Anatoly is proposing? [10:37] DougFoxvog: @Amanda: +1 [10:37] BobSchloss: @Cory - I don,t understand how Anatoly preserve coherence among the data mappings, the process mappings, the service requestor to service provider mappings... when all components of the distributed system are being designed "at the same time". What I am saying is that few big systems have all parts designed, specified, modeled AT THE SAME using the SAME REPRESENTATIONS. [10:38] BobSchloss: @Cory - sorry, I meant to write "when all the components of the distributed system are NOT being designed ,at the same time, [10:39] BobSchloss: I am going to try to find some time to read some tutorial about ISO 15926 (which I had not heard about before today) [10:40] CoryCasanave: @Bob, highly agile and "at the same time" can look similar [10:40] CoryCasanave: I also did not know about 15926 prior to speaking with Anatoly - must be a secret standard! [10:42] AmandaVizedom: It bears thinking about the extent to which federated systems come along with federated ontology development. That is, need not have (and may not be able to have) a central team doing all ontology development. You might have a central team doing some alignment, including mapping the points at which different domain ontologies and subsystem models connect with each other (and how; not just "sameAs" or "equivalentConcept" relationships), and where some upper ontology is needed or useful to support those connections. And it can work well to have some kinds of ontology done by professional ontologists, some done by various kinds of local (domain, data, process) exports. There is much research on where the sweet spots are in those divisions of labor (see, for example, DARPA,s Rapid Knowledge Formation (RKF) research program in the early 2000s). Newer folks to ontologies and semantic systems tend not to know about resea! rch like this, unfortunately. [10:43] NicolaGuarino: @Bob: be aware that ISO 15926 is explicitly *counterintuitive*. It aims at capturing "hard" engineering *facts*, so to speak, but not the everyday engineering talk, I would say... [10:44] PeterYim: == DennisWisknosky presenting ... [10:44] SimonSpero: BS on 15926: Against Idiosyncrasy in Ontology Development [ http://wings.buffalo.edu/philosophy/ontology/bfo/west.pdf ] [10:45] PeterYim: DennisWisnosky: EIW (semantic web project) demo will be available as of Mar-30 ... would like to show that to us [10:47] BobbinTeegarden: @Dennis BPMN not DoDAF up front? [10:47] PeterYim: DennisWisnosky: BLUF - "Bottom Line Up Front" [10:48] Tommi Karhela: @Nicola - Yes, those are usually first priciple mathematical models. However the equations are not described using ontologies. The configuration of the models (structure and parameters) are modeled using semantic graph and then these structures are used to initialize the actual solvers. [10:48] AnatolyLevenchuk: Tutorial about ISO 15926 -- http://fiatech.org/images/stories/techprojects/project_deliverables/iso-intro-ver1.pdf [10:50] MatthewWest: I presented an introduction to ISO 15926 in 2006. The slide set and podcast can be found here: http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2006_02_23 [10:51] NicolaGuarino: Unfortunately I have to go, sorry... [10:51] AnatolyLevenchuk: @SimonSpero: there was answer of MatthewWest to this criticism. I consider this criticism outdated and not so important. [10:52] BobSchloss: Dennis says (slide 3) H2R - hire to retire ontology; P2P - procure to pay ontology [10:54] AnatolyLevenchuk: If you want to try ISO 15926 you can take .15926 Editor here: http://techinvestlab.ru/dot15926Editor [10:56] MikeBennett: @Denis By BPMN Ontology, do you mean that the constructs in BPMN such as activities are defined as "Things" in ontology terms? We did something similar but I wasn,t sure if people would like it. [10:58] MatthewWest: Here is a link to a presentation I made to the 2009 Ontology Summit on IRING, which shows teh implementation architecture being developed for ISO 15926: http://ontolog.cim3.net/file/work/OntologySummit2009/OntologySummit2009_Symposium_20090406-07/The-iRING--Ontology-Summit--MatthewWest_20090406.ppt [10:58] MikeBennett: I have to jump off now for another call. [10:58] BobSchloss: Question for Dennis: Are the SPARQLMotion process/pipeline scripts generated by BPMN by software? by a develper using software? manually? [10:59] JamesOdell: @Mike - Similar to what I was alluding to in Anatoly,s presentation -- just not restricted to BPMN. [11:00] MikeBennett: @james thanks. I would like to explore this further in the FIBO work when we get to a suitable point. [11:00] BobSchloss: Dennis says: Once this is released to the people who build IT systems for the DoD organizations, they will have to teach people the proper way to model in this framework. [11:01] PeterYim: == CoryCasanave presenting ... [11:01] LeoObrst: @Dennis: How will you handle the switchover in the future to the ontology-based system you are building now from the existing database-based business and HR systems? Will you keep both active until people pound on the new system and get used to it? I assume you,ll be doing performance studies, etc., prior to the cutover? [11:02] AmandaVizedom: @MikeBennett - On slide 12, "Example Class Mappings", you can see in the screen shot that indeed, the BPMN constructs are defined in the bpmn: ontology. [11:02] JamesOdell: @Mike - Would love to discuss it with you. (Reston) [11:03] Tommi Karhela: @Dennis: Do you also simulate your business processes? We have been using system dynamics in Simantics to model business processes https://www.simantics.org/end_user_wiki/index.php/Simantics_System_Dynamics It would be interesting to try to transform BPMN based models into system dynamics simulation models in the same way that we are tranforming engineering CAD models into process simulation models. [11:04] MikeBennett: @James definitely. [11:05] SimonSpero: @Leo: The Spry prototype has been using Sparqlizers over RDBMS - http://www.information-age.com/channels/information-management/it-case-studies/1678933/qanda-us-dept-of-defense.thtml [11:06] AnatolyLevenchuk: @NicolaGuarino: situational method engineering have several standards now. I prefer ISO 24744 (metamodel for a development process) but there exist also OMG SPEM 2.0 [11:06] MatthewWest: My response to BS on ISO 15926: http://www.matthew-west.org.uk/documents/Reponse%20to%20Barry%20Smith%20Comments%20on%20ISO%2015926.htm [11:09] AmandaVizedom: Cory commented (while on slide 5) that while some fundamentals might be needed, the approach can,t require deep semantic or logical sophistication on the part of the (hundreds? or tens?) of thousands of people while will be producing the models.. [11:10] AliHashemi: I,ve never understood how shallow semantics can appropriately address systems federation or integration. Don,t you need to explicate on some level the ,deep, semantics to know whether the intent of a term or message will be preserved / honoured? [11:11] KenAllgood: @Amanda: that would seem to be a reasonable proposal.. Otherwise the adoption of the approach continues to be limited by the "entry cost" [11:11] SimonSpero: @Amanda @Ali: How deep is shallow [11:12] SimonSpero: @Simon: As shallow as possible, but now shallower [11:12] AliHashemi: You need precisely the amount of semantics that corresponds how the various systems treat the messages being passed... It seems like a somewhat arbitrary distinction (at least for interoperability) [11:12] AliHashemi: corresponds to* [11:12] AmandaVizedom: @Cory - I absolutely agree with that (your remark that I quoted above). It requires a well thought out distribution of the modelling work, and that has several components, IMHO. There is an interdependence between the the tools available for non-ontologists to work with and the sophistication they need to have. And to scale with the work needed, much of the modelling has to be done/doable at the local domain/expert/source level. [11:14] MatthewWest: Sadly I have to leave now [11:15] KenAllgood: @ Cory: Excellent breakdown of a reasonable and pragmatic approach.. [11:15] AmandaVizedom: ...@Cory, IME, a big chunk of domain experts, struggles with ontology work actually comes from badly aimed simplification -- even a day of conceptual fundamentals can make it so much clearer what is expected, what the model will do, etc. And those are fundamentals that most SMEs can get. [11:17] LeoObrst: @Ali: I agree. If you don,t capture the differences between the semantics of different communities, applications, etc., and instead jam them into common concepts, you really won,t get compliance. You,ll get lip-service and foot-dragging. Why? Because their real semantic distinctions are not honored. [11:17] SimonSpero: @Ali: I think that there,s a range of depths - some may only be enough for a subset of applications, but the cognitive load is critical [11:18] SimonSpero: @Ali, @Leo: The key is knowing what applications you can and can,t handle [11:18] BruceBray: @Amanda, do you have an example of what that "fundementals curriculum" looks like? - slide deck, course syllabus? [11:18] PeterYim: == Q&A and open discussion ... [11:19] AliHashemi: @Simon, Leo - I would think you need at least enough explicit semantics to ensure that a de-contextualized message preserves its intended function and doesn,t elicit unintended behaviours. Any less than that which would uncover important differences in the background assumptions would lead to the lip-service that Leo mentioned. I,m not sure how much you can confidently ascertain a priori, that the implicit background knowledge in each context / system doesn,t require additional "depth"... [11:20] KenAllgood: @Simon: Agree, it,s all about targeting the federation candidates.. Not attempting to solve world hunger per se. [11:21] AmandaVizedom: @Bruce: Not ready-made, as the examples, and parts, that I,ve done before have all been property of my employer at those times. However, I,ve been trying to think about how I might get time/opportunity to put something together that would be broadly useful & that I could put out there. [11:22] LeoObrst: @Simon: sure, I understand the architecture, but this goes to an issue John Sowa is always raising (and enables him to denigrate semantic web technology), i.e., that there is no gauge of performance differences prior to switchover to an OWL/RDF system, e.g.. [11:22] AliHashemi: @Simon - yes, it,s in the service of and scoped by the target applications (intended functions and behaviours). [11:22] HensonGraves: @ali,@leo, I disagree. one can not know the details of another person,s ontology, nor do we need this to have conversation. We can agree to meet on April 12 2012 at NIST without really understanding how we really understand NIST etc. [11:23] BobbinTeegarden: @Dennis Is there a map in the large to where in DoDAF all the parts fit? Can you share it? [11:24] SimonSpero: @Leo: [the sparqlizer performance issue?] I have asked about this at earlier pod demos. [11:24] AliHashemi: @Henson - I,m not sure the analogy holds. For one, we,ll be able to clear up any lingering confusion when we meet in person. A system can only process the given message. I do agree that the degree of agreement is correlated with the intended functionality of the message elements within the target system. But a blanket assertion that shallow semantics is adequate for federation is imo inaccurate. [11:25] SimonSpero: @Leo: There,s also a host of federation / linked data performance questions that need field testing [11:26] LeoObrst: @Henson: it,s because we already share a common ontology about NIST in quite great detail, as humans in USA. Do we need to have every filigree of knowledge in common? No, but the shared dense commonality is probably 95%, and the filagrees can be resolved in conversation. E.g., oh, you meant the NIST main campus in Gaithersburg. Which gate? Etc. [11:27] KenAllgood: Have to drop for another meeting.. Thanks All! [11:27] LeoObrst: @Bob: can you add your question to the chat room? I,m not sure I caught it. Thanks! [11:29] BobSchloss: My first question for Dennis was whether he intends to represent the status of all potential (already modeled business processes in BPMN) and all current "in flight processes" within an RDF triplestore, as well as to use the RDF triplestore to hold the information which different steps in the BPMN process produce, consume or test. He answered "Yes, he will do both". [11:30] LeoObrst: @Henson: also, just because we have common knowledge does not mean that we know what we know, i.e., we may not actually have that meta-knowledge. If I ask you how do you understand natural language, you may not know how to answer that, even though you are perfectly fluent. Why? Because you have not reflected, perhaps scientifically, on the kinds of knowledge you know about natural language understanding and its production. [11:30] SimonSpero: @Leo: a friend just deployed with the 2nd/5th Joked about the lag from the sat comms, but latency can be real problem with federating data over latency gaps [11:31] BobSchloss: My second question for Dennis was how is the translation from BPMN or BPMN+Rules into a process scripting language, like SPARQLMotion, done. He said they would like to always have this done by a software program rewriting into SPARQLMotion; not sure if that will always be able to be done with no human guidance; but so far, for their simple use cases, the translation has been able to be done automatically. [11:31] AmandaVizedom: @Ali - I completely agree. In general, the depth needed varies by example. Caveats: there are two different issues here: (1) the depth of semantics needed for a project overall, and (2) the depth of semantics that needs to be handled by the domain expert modelers who will likely need to do much of the effort. With that in mind, I,d suggest the re:(2) there is prior research that suggests types of ontology population that domain experts do well with a small amount of ontologist support, vs. areas where that never seems to work. And respect to (1) In a large-scale, federation contexts, it,s very likely that you can do local modeling with shallow semantics, but alignment requires deep semantics. And a common mistake is to think that those deep semantics needed for alignment can be an after thought. That depends on the methodology and representation you,ve chosen. If you,ve chosen an approach that doesn,t allow, or ! makes very burdensome, things like rules and second order classes and contextualization and so on, it will be much, much harder to do the alignment. [11:32] LeoObrst: @Bob: thanks! [11:32] Tommi Karhela: Unfortunately, I have to leave [11:33] AliHashemi: @Leo & Henson - It,s also worth noting that agreeing to meet at NIST requires agreement about what it is to actually meet at a place at a given time. For humans, as Leo notes, most of this is already common, but there is still ambiguity as to where in NIST, what time, how long for, etc. @Amanda - wholly agree. [11:35] Bob Smith 1 morphed into Bob Smith [11:37] BobSchloss: Leo verbally asks Dennis how the migration from the legacy applications (presumably do not have SPARQL interfaces) to new systems will be handled? Dennis answers that the driving case is Business Intelligence; this isn,t about replacing transactional application. [11:37] CoryCasanave: @Ali, this is where we need usable reference ontologies that have these base concepts worked out. Modeling then becomes more a practive of USING and COmbining concepts [11:37] BobSchloss: Dennis continues that legacy system will probably continue to be the system of record (my phrase) for next years. [11:38] BobSchloss: Dennis points out that different DoD services are evolving into a Cloud implementation with finer grained "services", replacing the current "21 big monolithic" applications... (Hope I got that right...) [11:40] BobSchloss: Fascinating session... I must leave now. Thanks to the presenters and the people contributing to the chat. [11:40] LeoObrst: @Peter: thanks, forgot! [11:40] CoryCasanave: Thanks for contributing @Bob [11:41] LeoObrst: Must go now too. Thanks for a very interesting session, all! [11:42] BobbinTeegarden: @Cory agree, probably no uberonto, but some set of common ontologies seem to be self organizing simply by usage. Since the web is so dynamic, this will probably continue to selforganize over time. [11:42] JoelBender: (thank you everyone) [11:42] PeterYim: great session ... thanks everyone! [11:42] SimonSpero: Good session [11:42] SteveRay: Very interesting [11:43] AmandaVizedom: Thanks, all, great food for thought! [11:43] PeterYim: -- session ended: 11:43am PST -- [11:43] AnatolyLevenchuk: Thank you all for a great session! [11:44] Sent transcript to: ailev@asmp.msk.su [11:44] Sent transcript to: cory-c@modeldriven.com ---------------