ppy/chat-transcript_unedited_20120216a.txt -------------- Chat transcript from room: summit_20120216 2012-02-16 GMT-08:00 -------------- [09:13] PeterYim: Welcome to the = OntologySummit2012: Session-06, Thursday 2012-02-16 = Summit Theme: OntologySummit2012: "Ontology for Big Systems" Track (4) Title: Large-Scale Domain Applications Session Topic: Large-Scale Domain Applications - Energy, Government and Geography Session Chairs: Dr. SteveRay (CMU) and Dr. TrishWhetzel (NCBO; Stanford) Panelists: * Dr. AndrewCrapo (General Electric) - "Overcoming Challenges Using the CIM as a Semantic Model for Energy Applications" * Dr. KrzysztofJanowicz (UCSB) - "Data-Intensive Geospatial Semantics" * Mr. BruceBauman (DoD) - "Separating Semantics and Implementation: From a Single Ontologically Sound Conceptual Model to Multiple Physical Schema Languages" * Mr. MillsDavis (Project10X) - "Enterprise Knowledge Computing - Case Examples" Session page: http://ontolog.cim3.net/cgi-bin/wiki.pl?ConferenceCall_2012_02_16 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: == [09:27] anonymous morphed into PaulWitherell [09:28] SteveRay: John Bilmanis, could you please change your chat room name to JohnBilmanis, which will then get properly linked to your ontolog forum namesake page. [09:29] anonymous1 morphed into Andrew Crapo [09:30] anonymous morphed into ChristopherSpottiswoode [09:31] Andrew Crapo morphed into AndrewCrapo [09:31] anonymous morphed into KenAllgood [09:31] anonymous1 morphed into BosseAndersson [09:31] John Bilmanis morphed into JohnBilmanis [09:31] James Odell morphed into JamesOdell [09:31] anonymous morphed into TomTinsley [09:32] anonymous1 morphed into GiancarloGuizzardi [09:33] anonymous morphed into hasan sayani [09:33] anonymous morphed into Bruce Bauman [09:34] anonymous morphed into TrishWhetzel [09:35] Bruce Bauman morphed into BruceBauman [09:35] anonymous morphed into Jacob Molenaar [09:37] anonymous1 morphed into NicolaGuarino [09:37] anonymous morphed into PeterDenno [09:43] TrishWhetzel: sorry, the call was dropped. [09:43] TrishWhetzel: let's move into the talks for now [09:43] anonymous morphed into LarryLefkowitz [09:47] ToddSchneider: What does IEC stand for? [09:48] SteveRay: International Electrotechnical Commission. One of the "big three" standards bodies (ISO, IEC, ITU) [09:53] DougFoxvog: There is no link to the paper underlined on page 5 [09:55] TerryLongstreth: slide 5: how do you know in advance how to order clauses to favor data more likely to need matching? [09:56] anonymous morphed into FrankOlken [09:56] JoelBender: As a big fan of OPS-5 back in the 80's, I would like to see how rules in Ret?? networks would assist in rule ordering. [09:57] TerryLongstreth: Never mind: you're reordering the query [09:57] SteveRay: Link to paper by Crapo: http://www.pointview.com/data/files/3/2433/1730.pdf [09:58] LeoObrst: @Andrew: most DBMSs and many Semantic Web reasoners will do query optimization as the first step in executing the query, and transform it. Really, that is always needed for efficiency. [10:00] anonymous1 morphed into ElizabethF [10:01] DougFoxvog: Optimization may lead to problems. E.g., if someone extends a geometry system to include spheres and cones, then a calculation of the area of an object as pi*radius^2 would be wrong, whereas if the type checking had not been optimized out, it would not be [10:02] TerryLongstreth: Being aware of the Open world assumption doesn't mean you can ignore it [10:03] PeterYim: == AndrewCrapo presenting ... (above) [10:04] LeoObrst: @Doug: yes, you can't change the domain ontology incompletely without also changing the optimization. [10:04] PeterYim: @ALL: the slides for MillsDavis has been uploaded, please refresh browser so it'll show up on your session page [10:06] anonymous morphed into Shirley Tseng [10:06] PeterYim: == KrzysztofJanowicz presenting ... [10:08] SimonSpero: @AndrewCrapo: question for later is how performance compares to some of the more sophisticated policy based dedicated routing algorithms; it's hard to figure the number of entities from the number of triples [10:08] Gary Berg-Cross: Andrew, what do use for topological and geometrical ontologies? [10:09] LeoObrst: @Andrew: yes, one can translate OWL with Open World Assumption into a representation that uses Closed World Assumption, and many folks do, at least locally. E.g., into logic programming (which has CWA). Using Answer Set Programming, one can interleave logical negation (OWA) and finite-failure-negation (CWA). Other work is addressing a more general solution for the Semantic Web. [10:09] AndrewCrapo: We didn't compare with policy based routing algorithms--we were looking at what could be done with OWL. As I recall the feeder network had ~50,000 nodes. [10:11] AndrewCrapo: The CIM defined the topological structure. We did not do GIS, if that's what you omean by geometrical. [10:11] CoryCasanave: @Andrew, we have done a lot of UML<>OWL translations - it sounds like you just have a bad transform spec, many transforms require a tweek for a particular UML style. So did you fix the transform or manualy fix the result? Let us know if you would like a better transform that has the features you mentioned. [10:12] Gary Berg-Cross: Andrew, by geometrical I was referring back to your mentioning circles with a radius etc. [10:12] AndrewCrapo: We used the CIM models supplied to us. I would be interested in better translators, definitely. [10:15] anonymous morphed into Yu Lin [10:15] BobbinTeegarden: @Andrew Did you do any cleanup or rationalization of the CIM model beforehand? [10:16] AndrewCrapo: GE Energy did quite a bit, otherwise they would have been totally unusable. We did a little, but that wasn't the focus of our effort. [10:21] AliHashemi: Sorry, I may have missed this, what did the ">" signify on slide 11? [10:21] NicolaGuarino: @Krzysztof: suppose that two different people, located in different geographical regions, tag the spot where they are as a "coffeeshop", while having different understandings of what a coffeeshop is (say, selling raw coffee vs. being a bar). How do you reconcile this heterogeneity using your method? [10:22] LeoObrst: @Ali: he skipped slide 11. [10:23] AliHashemi: (@Leo - Are you sure? I'm pretty confident he suggested that the picture on the upper left was of mountains, while the lower left was of cities.) [10:25] LeoObrst: @Ali: I thought that comment was for slide 10. [10:25] GiancarloGuizzardi: @Krzysztof: what do you mean by ontology design patterns here? [10:27] anonymous morphed into KarlG [10:29] KarlG morphed into KarlGrossner [10:29] SimonSpero: @SteveRay: That was sort of what I wondering when I asked about comparisons with dedicated routing systems. It's an argument for having a variety of strategies available within a reasoning system, and selecting them based on suitability and cost [10:31] PeterYim: == BruceBauman presenting ... [10:31] KrzysztofJanowicz: @Giancarlo: I mean strategies, you may also call them building blocks that domain experts can use to construct application ontologies. such patterns could include a path-movement patter, an event-activity-change pattern, a place pattern (those have been developed during the mentioned geovocamp) [10:32] GiancarloGuizzardi: @Krzysztof: Sounds very interesting. Can you provide some reference to those? [10:33] PeterYim: @KarlG ... can you let us know your full name, please (kindly morph to that by pressing the "Settings" button at top-center of the window) [10:33] AliHashemi: @Krzysztof - what did the ">" signify on slide 11? [10:34] KrzysztofJanowicz: topicX is more representative than topicY for a specific feature type [10:34] PeterYim: @KarlGrossner - thank you [10:35] KrzysztofJanowicz: Giancarlo, yes - I will send you an email; see also http://vocamp.org/wiki/GeoVoCampSB2012 [10:35] GiancarloGuizzardi: @Krzysztof: great thanks. [10:35] CoryCasanave: Bruce characterized OWL as an implementation language, not a conceptual modeling language. I tend to agree. What do people think about that? [10:36] ToddSchneider: Cory, yes. UML has too many embedded and implicit assumptions. [10:36] TerryLongstreth: @Cory: It depends on what you use it for. [10:37] CoryCasanave: If it depends on what you are using it for is it still a conceptual model? [10:39] TerryLongstreth: If I use Fortran to model a function, is it not a model? [10:40] CoryCasanave: @Terry, I didn't say it was not a mode, I said it is not a conceptual model of a domain. I don't think fortran is a conceptual model of a domain (other than the domain of computation) [10:40] TerryLongstreth: We're probably arguing about the term Conceptual. At the highest levels of abstraction, you probably wouldn't be able to automate interpretation of OWL [10:40] TerryLongstreth: Or better stated, you wouldn [10:41] TerryLongstreth: wouldn't find it useful. [10:42] CoryCasanave: So you don't think a conceptual model is useful? [10:42] TerryLongstreth: Not in anything other than natural language [10:44] GiancarloGuizzardi: Yes. Bruce is completely right. UML was just a pragmatic choice. It could have been (and be) anything else [10:44] Mills1: How are the rules (business logic of processes) defined? [10:44] CoryCasanave: @Terry, very much disagree conceptual models with an ontological grounding are essential for federating operational model, which is what Bruce is showing now. [10:45] TerryLongstreth: I was once on a task force to find a replacement for a proprietary design language. The proposed replacement was Pascal. The argument in its favor was the availability of syntax checkers. So the design language became implementable. When OWL is used for conceptual models, much of the conceptual weight is included as comments: NL. [10:46] ToddSchneider: What slide should we be on? [10:46] AliHashemi: 8 [10:47] GiancarloGuizzardi: 9, actually [10:47] AliHashemi: tx [10:47] ToddSchneider: Ali, I think he moved ahead [10:47] BobbinTeegarden: It would seem that in modeling, visual pictures count. Better grokking factor, and stakeholders lean in and try to fix things faster. Conceptual is a continuum, and yes fortran, text, UML 'pictures' -- all 'conceptual models' on a level [10:48] ToddSchneider: Graphics/images have a higher information density [10:53] TerryLongstreth: @Bobbin: better put than my attempt. Thanks. As soon as you "code" a concept for machine interpretation (whether it's OWL, CLIF, or UML), your model is an implementable (if it's syntactically correct). [10:54] GiancarloGuizzardi: my connection dropped for a while...please in which slide is Bruce in? [10:54] TrishWhetzel: Culture issues - are there any common themes that have made it easier or more difficult across projects? [10:55] TerryLongstreth: In choosing a language, you are restricting the range of expressiveness of concepts. [10:55] SteveRay: Slide 19 [10:55] BobbinTeegarden: Visualization is so key, and as Bruce is saying ... so missing. [10:55] GiancarloGuizzardi: Thanks a lot [10:55] BobbinTeegarden: @Terry Yes. [10:56] TrishWhetzel: Talent issue - what training programs would you suggest to address this issue? [10:57] NicolaGuarino: @Bruce: what is the most important advantage of OntoUML, for your applications? [10:57] PeterYim: @Giancarlo - Bruce (on slide#18) was commenting on the fact about the lack of tools to flip back-and-forth between the UML models and the semantic model ... any plans on your part (or you know of) on the development of those badly needed tools? [10:57] Mills1: un mute everyone for a moment [10:58] CoryCasanave: @Terry (Bruce could comment), the essential difference is what you are modeling. If you are modeling an interpretation of "the world" that is nto the same as a model of information about that world or algorithums that may perform some function in it. [10:58] PeterYim: == MillsDavis presenting ... [10:59] CoryCasanave: @Peter, re: "tools to flip back-and-forth between the UML models and the semantic model " That is part of the SIMF effort I have mentioned. [10:59] GiancarloGuizzardi: @Peter: great question. We have been developing a modeling environment which deals with aspects such as modeling with automatic verification support, validation support via visual simulation and mapping to different codification languages (OWL included, with different mappings with different expressivity) [11:00] GiancarloGuizzardi: However, what we have now is an academic prototype. [11:00] BobbinTeegarden: @Giancarlo Any chance of a visual diff? [11:01] GiancarloGuizzardi: @Bobbin: Could you please elaborate on the question? [11:01] AliHashemi: @Hand: refresh session page for slides [11:01] ToddSchneider: Giancarlo, any chance to play with your prototype? [11:02] GiancarloGuizzardi: Of course. We have implemented a the full metamodel of the language and the associated OCL constraints (that restrict the syntactical models to those which are ontologically correct - according to the underlying theory) [11:02] GiancarloGuizzardi: that is fully available [11:03] BobbinTeegarden: @Giancarlo E.G. see two ontologies at the same time and highlight the differences (like Linux diff) [11:03] SimonSpero: @bobbin: owldiff? [11:03] GiancarloGuizzardi: one of the prototypes can be found here: http://code.google.com/p/rcarraretto/ [11:04] ToddSchneider: Giancarlo, Excellent. Thank you. [11:04] GiancarloGuizzardi: another one can be found here: http://code.google.com/p/ontouml/ [11:04] SimonSpero: @Bobbin; http://sourceforge.net/projects/owldiff/ [11:05] GiancarloGuizzardi: We are very much willing to collaborate with people on advancing these tools, btw [11:05] SteveRay: We are on slide 4 [11:05] PeterYim: @Giancarlo, @Cory - just to supplement the earlier question - BruceBauman was specifically referring to the visualization tools "to flip back-and-forth between the UML models and the semantic model" [11:05] GiancarloGuizzardi: @Bobbin: That would be an interesting feature. The prototypes that we have do not have that feature. [11:06] GiancarloGuizzardi: @Peter: that is a tool that Bruce developed with his team [11:07] SimonSpero: I wonder if you could encode theBritish Nationality Act as a logic program? http://www.doc.ic.ac.uk/~rak/papers/British%20Nationality%20Act.pdf [11:07] GiancarloGuizzardi: @Todd: we are developing a better tool now with most of the features which were developed over the years. If you are interested in playing with that, please send me an email (we dont have a weblink to that one yet...) [11:07] PeterYim: @Giancarlo - BruceBauman should jump in here ... I believe in his presentation, he said this is not something they wanted to develop (if I heard him correctly) [11:07] PeterYim: @11:07am PST - now on MillsDavis' slide#5 [11:08] GiancarloGuizzardi: @Peter: one of the prototypes we have developed do include a transformation to several different versions of OWL. It does not provide any Round-Trip Engineering feature yet. [11:08] CoryCasanave: @Andrew, here is a link to the open SW we did that translates between OWL, UML and some others, but it is old, the newer versions are not published yet: http://portal.modeldriven.org/project/EKB [11:10] PeterYim: @Giancarlo - thanks [11:10] GiancarloGuizzardi: By different versions of OWL, I mean different codifications capturing different design decisions and which capture different "portions" of the modal semantics of the original language [11:11] SteveRay: @Mills: Did your cases try to migrate from any legacy models, or go completely greenfield? [11:13] SteveRay: Question for all the speakers: Where in the systems engineering cycle was ontology most successfully applied? [11:13] NicolaGuarino: @Mills: you show the importance of model-based systems in public administration. But how to make sure you get these models right? Where are the problems? [11:15] CoryCasanave: @miils: is this an ontology at runtime or ontology produces runtime approach? [11:15] GiancarloGuizzardi: @Todd: I have just noticed that the two links I sent you do not have a clear pointer to the fuller reference to OntoUML and (more importantly) the ontological theory behind it which is: http://www.inf.ufes.br/~gguizzardi/OFSCM.pdf [11:16] AmandaVizedom: @Mills: What approaches were used to verify that the semantic models captured the correct knowledge? Were ontologies built for the task, reused, or both? Was evaluation done on semantic models, on whole system, or both? [11:17] ToddSchneider: Giancarlo, thank you. [11:17] Mills1: Modeling in the 4 cases addresses all aspects -- data, process, decision, user experience, and infrastructure (legacy and external) [11:18] CoryCasanave: General question (Related to question to Mills): Do always, sometimes or never want the ontology and the runtime to be the same thing? Runtime systems have some different concerns, where are those concerns injected? [11:19] ToddSchneider: Giancarlo, I didn't recognize the file name. I already have it. In fact I just printed it out yesterday to make mediation easier. [11:20] ToddSchneider: Mills, w.r.t. to systems engineering processes, does the modeling you were describing include requirements, architecture, design, test, ... [11:21] GiancarloGuizzardi: @Todd: please, any question, comment or opportunity for collaboration is very welcome. [11:21] GiancarloGuizzardi: @todd: btw, the (very sensible) simplification of the language presented by Bruce can be found in http://www.omgwiki.org/architecture-ecosystem/doku.php?id=uml_based_data_modeling_for_an_enterprise_data_model [11:22] ToddSchneider: Giancarlo, questions? I always have questions. Time for finding answers is another question. [11:22] Mills1: Cory, the separation of concerns between the business logic and the system behaviors is where important IT architectural decisions come into play. For example, if your data modeling exposes many sources, kinds of information and can map them to a graph in the middle, this still begs the question of query performance -- should the data be ETL and in one place, or left to be accessed via federation? Also, the notion of parallelization of the reasoning across multiple engines. [11:23] GiancarloGuizzardi: @Todd: please send them to me...I will do my best to reply to them a.s.a.p. if I can help. [11:23] NicolaGuarino: @Giancarlo&Bruce: I don't like very much the terminology used for different kinds of identity in slide 8, "unified" vs "dispersive". I think I understand it, but it is confusing... [11:24] BruceBauman: Thats a fair comment, although I've struggled with coming up with better language. [11:25] BruceBauman: One of the challenges we face is communicting across different sub-disciplines such as data modeling, ontology, bussiness rules etc, acadamia etc. [11:25] GiancarloGuizzardi: @Nicola: I think what Bruce mean by unified is "having a common principle of identity shared by all the instances" [11:25] BruceBauman: There is no common set of terms. [11:25] GiancarloGuizzardi: if think to his audience is easier than Sortal [11:25] NicolaGuarino: @Giancarlo: yes, I got it :-) [11:26] GiancarloGuizzardi: @Nicola: dispersive types as a synonym for non-sortal was a term introduced by Eli Hisch (in the Concept of Identity) [11:26] GiancarloGuizzardi: Eli Hirsch [11:26] Mills1: Todd, yes. But it's not a waterfall. [11:27] NicolaGuarino: @Giancarlo: yes, but these terms remind me issues concerning unity, which is another story [11:27] GiancarloGuizzardi: @Nicola: I see you point. [11:27] GiancarloGuizzardi: your point [11:27] BobbinTeegarden: @Mills Are you hinting that the model IS the message? Any tools? [11:28] NicolaGuarino: Still I have a question for Bruce: what is the most practical advantage of using OntoUML? What is the argument you use with your people? [11:29] Mills1: Going back to Steve Ray's question about greenfield or adaptation of legacy. The answer is both. and Both with respect to data and systems process interoperability. [11:29] NicolaGuarino: Sorry I was disconnected [11:29] NicolaGuarino: No problem... [11:30] PeterYim: @Nicola - Bruce is responding (verbally) ... you'll have to catch that in the recording later. [11:30] Mills1: Bruce -- What about the business logic in relation to decisions and processes? [11:32] Mills1: I will write a response [11:32] LeoObrst: Good session. Goodbye! [11:33] TrishWhetzel: Thanks Steve! [11:33] AmandaVizedom: @Mills - Great, thanks. [11:33] JoelBender: Thank you! [11:33] Mills1: sbye [11:33] AmandaVizedom: Great session, Thanks! [11:33] KrzysztofJanowicz: bye bye [11:33] NicolaGuarino: bye [11:33] PeterYim: great session! ... I'll leave the chat session on for another 10 minutes [11:33] PeterYim: -- session ended: 11:32 am PST -- [11:33] SteveRay: Thanks, Peter [11:33] BruceBauman: Nicola, OntoUML provides three benifits. [11:34] BruceBauman: 1. It is expressive enough so that we can always come up with a pattern to convert from it to any given implmentation. [11:34] BruceBauman: 2. It has bulit in rules that allow us to catch early some modeling errors and clarify what semantic meaning is intended. [11:35] BruceBauman: 3. It is flexible enough to allow us to effectivly capture and relate different projects models, as well as precise enough in its us of the constructs of the langauge to minimize modeling "construct variability" [11:36] NicolaGuarino: Very good, Bruce. I think this is very useful. Nice talking to you, bye [11:36] BruceBauman: It was a pleasure. [11:38] JoelBender: How much of the underling Alloy text is exposed to the UnoUML user? [11:41] JoelBender: While I appreciate that "pure" NL isn't succinct enough, I'd be reluctant to hoist Alloy on to domain experts that aren't also fairly experienced programmers. [11:49] JoelBender: Maybe a quick class in Clojure before introducing statements like 'fact { eats = Fox->Chicken + Chicken->Grain}; :-) [11:50] JoelBender: Thank you to the presenters, and thanks again Peter! --------------