From OntologPSMW

Jump to: navigation, search
[ ]

Contents

Ontology Summit 2013: Panel Session-09 - Thu 2013-03-14     (1)

Summit Theme: "Ontology Evaluation Across the Ontology Lifecycle"     (1A)

Summit Track Title: Track-C: Building Ontologies to Meet Evaluation Criteria     (1B)

Session Topic: Ontology Development Methodologies for Reasoning Ontologies     (1C)

Session Co-chairs: Mr. MikeBennett (EDM Council; Hypercube, UK) and Dr. MatthewWest (Information Junction, UK) - intro slides     (1D)

Panelists / Briefings:     (1E)

  • Dr. JoanneLuciano (RPI-TWC, US) - "A Generalized Framework for Ontology Evaluation (GOEF)" slides     (1F)
  • Dr. LeoObrst (MITRE, US) - "Developing Quality Ontologies Used for Reasoning" slides     (1G)

Abstract     (1I)

OntologySummit2013 Session-09: "Ontology Development Methodologies for Reasoning Ontologies" - intro slides     (1I1)

This is our 9th Ontology Summit, a joint initiative by NIST, Ontolog, NCOR, NCBO, IAOA & NCO_NITRD with the support of our co-sponsors. The theme adopted for this Ontology Summit is: "Ontology Evaluation Across the Ontology Lifecycle."     (1I2)

Currently, there is no agreed methodology for development of ontologies, and there are no universally agreed metrics for ontology evaluation. At the same time, everybody agrees that there are a lot of badly engineered ontologies out there, thus people use -- at least implicitly -- some criteria for the evaluation of ontologies.     (1I3)

During this Ontology Summit, we seek to identify best practices for ontology development and evaluation. We will consider the entire lifecycle of an ontology -- from requirements gathering and analysis, through to design and implementation. In this endeavor, the Summit will seek collaboration with the software engineering and knowledge acquisition communities. Research in these fields has led to several mature models for the software lifecycle and the design of knowledge-based systems, and we expect that fruitful interaction among all participants will lead to a consensus for a methodology within ontological engineering. Following earlier Ontology Summit practice, the synthesized results of this season's discourse will be published as a Communique.     (1I4)

At the Launch Event on 17 Jan 2013, the organizing team provided an overview of the program, and how we will be framing the discourse around the theme of of this OntologySummit. Today's session is one of the events planned.     (1I5)

In this 9th virtual panel session of the Summit, we will look at methodologies for developing ontologies used for reasoning applications.     (1I6)

More details about this Ontology Summit is available at: OntologySummit2013 (homepage for this summit)     (1I7)

Briefings     (1I8)

  • Dr. JoanneLuciano (RPI-TWC, US) - "A Generalized Framework for Ontology Evaluation (GOEF)" slides     (1I8A)
    • Abstract: ... This talk willpresent an approach for a framework for Ontology evaluation and engineering. It presents the role of the use case for evaluating function, the standards requirements (business or external utility), and components, which are modules that can be individually validated. The approach can be utilized for modular development and test. It is suggested that a SADI service (sadiframework.org) would be a good implementation.     (1I8A1)
  • Dr. LeoObrst (MITRE, US) - "Developing Quality Ontologies Used for Reasoning" slides     (1I8B)
    • Abstract: ... Ontologies are by default intended for reasoning over. Otherwise simpler semantic models should be used. This talk will discuss ontology development practices that lead to ontologies having greater quality. Intrinsic and extrinsic properties apply, and apply across the ontology lifecycle, but the ontology development cycle must include a projected architecture as to how it will be used. The architecture includes what the reasoning component will consist of, and the language it will use.     (1I8B1)

Agenda     (1J)

OntologySummit2013 - Panel Session-09     (1J1)

  • Session Format: this is a virtual session conducted over an augmented conference call     (1J2)

Proceedings     (1K)

Please refer to the above     (1K1)

IM Chat Transcript captured during the session    (1K2)

see raw transcript here.     (1K2A)

(for better clarity, the version below is a re-organized and lightly edited chat-transcript.)     (1K2B)

Participants are welcome to make light edits to their own contributions as they see fit.     (1K2C)

-- begin in-session chat-transcript --     (1K2D)


Chat transcript from room: summit_20130314     (1K2E)

2013-03-14 GMT-08:00 [PDT]     (1K2F)


[10:10] Peter P. Yim: Welcome to the     (1K2G)

Ontology Summit 2013: Virtual Panel Session-09 - Thu 2013-03-14     (1K2H)

Summit Theme: Ontology Evaluation Across the Ontology Lifecycle     (1K2I)

  • Summit Track Title: Track-C: Building Ontologies to Meet Evaluation Criteria     (1K2J)

Session Topic: Ontology Development Methodologies for Reasoning Ontologies     (1K2K)

Mr. Mike Bennett (EDM Council; Hypercube, UK) and Dr. Matthew West (Information Junction, UK)     (1K2M)

Panelists / Briefings:     (1K2N)

Logistics:     (1K2Q)

  • (if you haven't already done so) please click on "settings" (top center) and morph from "anonymous" to your RealName (in WikiWord format)     (1K2S)
    • for Linux Skype users: please note that the dial-pad is only available on v4.1 (or later or the earlier Skype versions 2.x,) if the dialpad button is not shown in the call window you need to press the "d" hotkey to enable it.     (1K2W1)

Proceedings:     (1K2AD)

[10:19] anonymous morphed into AstridDuqueRamos     (1K2AE)

[10:23] anonymous morphed into Bob Schloss     (1K2AF)

[10:24] Astrid: Hello Peter     (1K2AG)

[10:24] Astrid: I am not able to speak     (1K2AH)

[10:24] Astrid: I have problems with my microphone     (1K2AI)

[10:24] Astrid: But I will listen.     (1K2AJ)

[10:25] Lamar Henderson morphed into Lamar Henderson     (1K2AK)

[10:25] anonymous morphed into Michel Dumontier     (1K2AL)

[10:26] Astrid morphed into Astrid Duque     (1K2AM)

[10:27] Peter P. Yim: No problem, Astrid ... I was suggesting that you (and everyone else who hasn't     (1K2AN)

already) to morph your name into the nominal WikiWord name (as per your identity on our wiki) - like     (1K2AO)

AstridDuqueRamos, or Joanne Luciano, Bob Smith, etc. ... thanks (that helps the automatic link     (1K2AP)

generation on the wiki)     (1K2AQ)

[10:29] Astrid Duque morphed into AstridDuqueRamos     (1K2AR)

[10:30] Peter P. Yim: Hi Michel, welcome!     (1K2AT)

[10:30] Steve Ray: Lots of hiss on the line.     (1K2AU)

[10:30] Bob Schloss: On the telecon audio bridge, there is a heavy buzzing which is making it hard     (1K2AV)

for me to hear you, Peter, as well as the other person who is speaking.     (1K2AW)

[10:30] Bob Schloss: If I am the only person who hears the hissing, I can call in again.     (1K2AX)

[10:30] Mike Bennett: We all hear it. Also someone talking.     (1K2AY)

[10:30] Mike Bennett: This will clear when Peter mutes everyone     (1K2AZ)

[10:31] Bob Schloss: It sounds like 2 telecon lines are crossed.....     (1K2AAA)

[10:31] anonymous morphed into Ali Hashemi     (1K2AAB)

[10:31] Amanda Vizedom: Yes, in the meantime, perhaps those who aren't speaking can self-mute (*6)     (1K2AAC)

[10:31] anonymous1 morphed into Torsten Hahmann     (1K2AAD)

[10:32] anonymous morphed into Doug Foxvog     (1K2AAE)

[10:32] anonymous1 morphed into Michael Denny     (1K2AAF)

[10:34] anonymous morphed into Ludger Jansen     (1K2AAG)

[10:34] Peter P. Yim: == Mike Bennett opening the session on behalf of the co-chairs ... ... see: the     (1K2AAH)

[0-Chair] slides     (1K2AAI)

[10:35] anonymous morphed into RosarioUcedaSosa     (1K2AAJ)

[10:37] Samir Tartir morphed into Samir Tartir     (1K2AAP)

[10:38] anonymous: where can one download the slides     (1K2AAQ)

[10:39] Peter P. Yim: links to slides can be found under:     (1K2AAR)

[10:41] Peter P. Yim: == Joanne Luciano presenting ... see: the [ 1-Luciano ] slides     (1K2AAT)

[10:48] GaryBergCross: Is the link for SADI service (sadi.org) on the page the correct???     (1K2AAU)

[11:00] Mike Dean: @Gary SADI service is at http://sadiframework.org     (1K2AAV)

[11:03] GaryBergCross: Mike, yes that is what I thought. The session page has it wrong.     (1K2AAW)

[11:03] Peter P. Yim: @GaryBergCross, @MikeDean - thank you, the link has been updated on the session     (1K2AAX)

[10:51] Terry Longstreth: Joanne - can you give an open reference /url for the Cockburn Method?     (1K2AAZ)

appears to be broken there --we have a local copy (and I'll send the to Alistair to let him know. I     (1K2AAAB)

think I forgot to do this when I was in contact with him a while ago:     (1K2AAAC)

http://tw.rpi.edu/media/latest/UseCase-Template_SeS (someone please try this RPI link and verify it     (1K2AAAD)

works). thanks     (1K2AAAE)

[11:05] Mike Bennett: That link works.     (1K2AAAF)

Microsoft Word template     (1K2AAAH)

[11:07] Joanne Luciano: @MikeBennett Thanks for checking.     (1K2AAAJ)

[11:23] Joanne Luciano: For those downloading the use case template -- recall that is the "starting     (1K2AAAK)

point" -- that is what needs to be formalized so that an ontology (or part of one) can be evaluated     (1K2AAAL)

with respect to it.     (1K2AAAM)

[10:55] anonymous morphed into Bobbin Teegarden     (1K2AAAN)

[10:48] Hans Polzer: Regarding Joanne's presentation, I would suggest looking at a range of contexts     (1K2AAAO)

for ontology evaluation, not just a use case. Another way to look at this is as a set of use cases     (1K2AAAP)

that span the range of ontology application contexts     (1K2AAAQ)

[10:51] Mike Bennett: @Hans I think that defines a difference between an ontology which is to be     (1K2AAAR)

developed for a specific application, and an ontology that is to be developed as a standard, which     (1K2AAAS)

would necessarily be use case agnostic. Though of course, if ontologies developed for applications     (1K2AAAT)

are designed to be broader than just one use case, then they can be reused by others. I think this     (1K2AAAU)

opens up an interesting line of discussion we can cover in the discussion.     (1K2AAAV)

[10:53] Matthew West: I know people who argue that Use Cases are inadequate as a statement of     (1K2AAAW)

requirements, because they are point requirements, whereas what you really want is to develop a     (1K2AAAX)

required capability. Use cases can be very helpful in informing such a required capability, and     (1K2AAAY)

later illustrating it, but are not a substitute.     (1K2AAAZ)

[10:54] Amanda Vizedom: @Hans - alternative but compatible view (I think): identify elements or     (1K2AAAAA)

characteristics of use cases that matter (or are hypothesized to matter) for ontology suitability,     (1K2AAAAB)

so that these elements can be reused to analyze and describe other use cases. Some of these will     (1K2AAAAC)

clearly be contextual, but the path to reuseful reasoning is better defined. As I understand it     (1K2AAAAD)

(Joanne can confirm/disconfirm) this is part of the first-stage support that is envisioned for GOEF.     (1K2AAAAE)

[10:56] Todd Schneider: Use cases can be used to discover requirements (i.e., an analysis tool).     (1K2AAAAF)

[10:57] Hans Polzer: Use cases also have lots of implicit scope, which can lead to overlooked     (1K2AAAAG)

requirement assumptions. On the other hand, a set of use cases can help understanding of a range of     (1K2AAAAH)

application contexts, since "range" is often difficult to conceptualize while use cases are     (1K2AAAAI)

typically presented as quasi-concrete examples.     (1K2AAAAJ)

[10:59] Amanda Vizedom: @Matthew: I would agree with that; there is an analysis stage between use     (1K2AAAAK)

case description and technical requirements specification that utilizes knowledge of both the     (1K2AAAAL)

operational background/need and the technical specifics.     (1K2AAAAM)

[10:59] Terry Longstreth: Joanne question - Does your methodology necessarily reveal the existence of     (1K2AAAAN)

an Ontology?     (1K2AAAAO)

[11:12] Joanne Luciano: @TerryLongstreth, I don't understand your question. Let's discuss during the     (1K2AAAAP)

discussion.     (1K2AAAAQ)

[11:00] Megan Katsumi: @Amanda: Can you clarify what you mean by elements or characteristics of use     (1K2AAAAR)

[11:06] Amanda Vizedom: @Megan - I should probably let Joanne talk about what she means first. ;-) At     (1K2AAAAT)

least, if I'm correct in saying that this is a reasonable way to describe what she is talking about     (1K2AAAAU)

in the description of the 1st stage of the GOEF framework, in which the use case is analyzed and     (1K2AAAAV)

specified. I think, from Joanne's description (and previous discussions), that this specification is     (1K2AAAAW)

not open-ended but focuses on capturing particular aspects of the use case that are known or     (1K2AAAAX)

believed to be needed as input in order to evaluate how well various candidate ontologies might fit     (1K2AAAAY)

that use case.     (1K2AAAAZ)

[11:06] Joanne Luciano: my original presentation from 2008 is on slideshare (you'll notice I used     (1K2AAAAAA)

Leo's template for that (I was at MITRE).     (1K2AAAAAB)

[11:50] Joanne Luciano: should it come up. here's the link to the slideshare presentation:     (1K2AAAAAC)

(from 2008!) it was a proposal for internal research at MITRE.     (1K2AAAAAE)

[11:07] Joanne Luciano: I am listening to Leo and I still have a bad cold, so I'm happy to discuss     (1K2AAAAAF)

after during discussion.     (1K2AAAAAG)

[11:00] Peter P. Yim: == Leo Obrst presenting ... see: the [ 2-Obrst ] slides     (1K2AAAAAH)

[11:03] Fabian Neuhaus: Leo: could you tell us when you switch slides?     (1K2AAAAAI)

[11:09] Doug Foxvog: There can be problems with a separation into T-Box & A-Box. In nature guides,     (1K2AAAAAJ)

for example, lots of statements are made of properties of classes (e.g., types of living thing).     (1K2AAAAAK)

These classes, themselves are instances of higher-order classes (Species, Genus, ... Phylum, etc.).     (1K2AAAAAL)

These types are both instances & classes. The same thing happens when talking about part types. The     (1K2AAAAAM)

classic Wine Ontology had huge problems in expressing properties of types of wine, ending up     (1K2AAAAAN)

defining the narrowest classes as instances of the next broader class, because of restrictions of     (1K2AAAAAO)

the A/T box restrictions it was operating under.     (1K2AAAAAP)

[11:10] Amanda Vizedom: @Megan, @Joanne, I'd expect, for example, that the kinds of things Leo is     (1K2AAAAAQ)

discussing now (slides 2-3: how complex is the reasoning needed for the use case, if any?) would be     (1K2AAAAAR)

a subgroup of the use case characteristics that can be described and matter for ontology     (1K2AAAAAS)

suitability.     (1K2AAAAAT)

[11:11] Doug Foxvog: Leo discussing DL not being able to represent many rule types. I suggest that     (1K2AAAAAU)

this is a reason to avoid using DL languages.     (1K2AAAAAV)

[11:12] Mike Bennett: @Doug avoid in general, or avoid for specific requirements / usages of the     (1K2AAAAAW)

[11:14] Amanda Vizedom: @doug +1 regarding the Wine Ontology example. It is a good example of the     (1K2AAAAAY)

approach suggested, and choices forced, by a particular kind of DL-based language. It is also a good     (1K2AAAAAZ)

example of why for most reasoning, or cross-specialty integration, applications, that approach is     (1K2AAAAAAA)

[11:14] Doug Foxvog: @Mike: I suggest avoiding unless the only use of the "ontology" is to be used as     (1K2AAAAAAC)

a taxonomy. If you use rules or want to integrate, a more powerful language is called for, imho.     (1K2AAAAAAD)

[11:15] Mike Bennett: @Doug - thanks. I guess that comes under the heading of what Leo calls "not an     (1K2AAAAAAE)

ontology" :)     (1K2AAAAAAF)

[11:17] anonymous morphed into Pavithra Kenjige     (1K2AAAAAAH)

[11:19] Doug Foxvog: Re Slide 7, "tractable reasoning": even a higher order language can exhibit     (1K2AAAAAAI)

tractable reasoning, when the rules reasoned over permit it. Note that normal computer languages are     (1K2AAAAAAJ)

intractable, but people still program using them. A programmer would not want to program in a     (1K2AAAAAAK)

guaranteed tractable language.     (1K2AAAAAAL)

[11:18] Pavithra Kenjige: Requirements are written in statements can be shown as use cases and     (1K2AAAAAAM)

scenarios to capture how user use a system to meet those requirements     (1K2AAAAAAN)

[11:20] Pavithra Kenjige: So in my opinion, Use Cases and scenarios and test cases that are based on     (1K2AAAAAAO)

Use Cases & scenarios are verification or evaluation of meeting the requirements     (1K2AAAAAAP)

[11:21] Joanne Luciano: I think SADI services can help - I'd like to incorporate GOEF development     (1K2AAAAAAQ)

into https://github.com/timrdf/DataFAQs/wiki I'd love to have some "clinics" to discuss and develop     (1K2AAAAAAR)

the implementation. I have a paper in the works that will use the iChoose example throughout. The     (1K2AAAAAAS)

current draft version doesn't have a single example (because these ideas have only been able to be     (1K2AAAAAAT)

developed in fits and starts). It includes some mock up screen shots to help get ideas across.     (1K2AAAAAAU)

RE: not an ontology -- the 'spectrum' seems to help cover bases and avoid "semantic" arguments     (1K2AAAAAAV)

(which is a challenge in a "semantic" community)     (1K2AAAAAAW)

[11:22] Doug Foxvog: Re slide 8. 3D/4D reasoning can be left out of most ontologies. If a given     (1K2AAAAAAX)

application requires committing to one or the other, it can also inherit a small ontology that     (1K2AAAAAAY)

specifies the 3D+1 or 4D theory.     (1K2AAAAAAZ)

[11:25] Doug Foxvog: re slide 9: There is a huge ontology of "part-of". Transitivity often won't work     (1K2AAAAAAAA)

if one switches between different types of "part-of".     (1K2AAAAAAAB)

[11:26] Mike Bennett: @Doug presumably part terms like "nearside front wheel" versus part terms like     (1K2AAAAAAAC)

Wheel have different requirements in this regard.     (1K2AAAAAAAD)

[11:29] Doug Foxvog: @Mike: Yes.     (1K2AAAAAAAE)

[11:29] GaryBergCross: @MikeBennett wheels may be "components" rather than a simple part.     (1K2AAAAAAAF)

[11:31] Mike Bennett: @Doug thought so. I think distinguishing between those usages (e.g. what is a     (1K2AAAAAAAG)

Component) per Gary, depends on using a suitable upper ontology that distinguishing     (1K2AAAAAAAH)

something-in-a-role from something-in-itself. So there is a dependency between these two     (1K2AAAAAAAI)

considerations I think.     (1K2AAAAAAAJ)

[11:31] Doug Foxvog: @Gary: "componentOf" would be a specialization of "physicalSubpartOf".     (1K2AAAAAAAK)

[11:33] GaryBergCross: I sometimes think of these bridging, integrative ontologies as simpler     (1K2AAAAAAAL)

bridging schemas.     (1K2AAAAAAAM)

[11:33] Doug Foxvog: @Mike: Yes. the truth of (componentOf Car2087532408 Wheel234752347) is time     (1K2AAAAAAAN)

[11:34] GaryBergCross: @Doug Yes that is a good specialization.     (1K2AAAAAAAP)

[11:29] Doug Foxvog: re slide 10: Do you need human capability to formulate queries using the     (1K2AAAAAAAQ)

computer's query language? A user interface should obviate this. One does not require a human     (1K2AAAAAAAR)

capability to formulate SQL queries for someone to ask for information from a database.     (1K2AAAAAAAS)

[11:33] anonymous morphed into Fred Hosea     (1K2AAAAAAAT)

[11:35] Doug Foxvog: Re slide 12: "Anyone we know" might be Insectivore.     (1K2AAAAAAAU)

[11:36] Doug Foxvog: The Venn diagram does not intersect Reptile and Mammal, so the "Anyone we know"     (1K2AAAAAAAV)

circle includes some mammals, some reptiles, and some other living things.     (1K2AAAAAAAW)

[11:35] Amanda Vizedom: Relating Leo's talk to my own (two weeks ago,)     (1K2AAAAAAAX)

most of the questions Leo is talking about are the sorts of things that should be asked     (1K2AAAAAAAZ)

at the stage when "business requirements" are being transformed into technical requirements     (1K2AAAAAAAAA)

(for the ontology and/or system). The answers should: figure into ontology/system design;     (1K2AAAAAAAAB)

be used to development constraints, goals, steps; be used to identify what should be evaluated     (1K2AAAAAAAAC)

when to make sure requirements are being met and maintained. I would say that this is the     (1K2AAAAAAAAD)

knowing-what-to-develop/knowing-what-to-check-for aspect of the Big Issues in Ontology     (1K2AAAAAAAAE)

Evaluation; the how-to-check-for-X/evaluation methods aspect is a parallel Big Issue.     (1K2AAAAAAAAF)

[11:38] Mike Bennett: @Amanda +1 - these kinds of technical considerations are a parallel to what in     (1K2AAAAAAAAG)

conventional technology development would be non functional requirements - but they clearly deliver     (1K2AAAAAAAAH)

benefits to the business integrity of a model of the business domain. That opens up a lot of     (1K2AAAAAAAAI)

interesting questions.     (1K2AAAAAAAAJ)

[11:39] Matthew West: @LeoObrst - please go ahead and cover you backup slides as well.     (1K2AAAAAAAAK)

[11:40] Amanda Vizedom: re: collaborative development: big tool need in the OWL world: tunable     (1K2AAAAAAAAL)

automatic checking on addition (commit-time; rejection if inconsistent). More comprehensive     (1K2AAAAAAAAM)

automatic bookkeeping on assertions and inferences so that problems can be debugged, fixed and     (1K2AAAAAAAAN)

modules/packages re-submitted.     (1K2AAAAAAAAO)

[11:40] Joanne Luciano: I'm wondering how many times Leo has been around the circles and items on     (1K2AAAAAAAAP)

slides#18-21. I don't think I want to count. BTW, Happy Pi Day (3.14 in the way the US does month     (1K2AAAAAAAAQ)

[11:46] Amanda Vizedom: @Leo - slide 20 - yes indeed, often looped around, etc., *and* often there     (1K2AAAAAAAAS)

are interdependencies with other components of a larger system. At various points, there is or     (1K2AAAAAAAAT)

should be cross-checking to see whether things are (still) in sync, whether they are on-track to     (1K2AAAAAAAAU)

work together as desired.     (1K2AAAAAAAAV)

[11:47] Mike Bennett: @Amanda +1 - not only for collaborative development but I think we need a tool     (1K2AAAAAAAAW)

that does for ontologies what UML does for software designs, including what you have there, and also     (1K2AAAAAAAAX)

visualizations both to business domain and to implementers.     (1K2AAAAAAAAY)

[11:48] Amanda Vizedom: @MikeBennett: Agreed.     (1K2AAAAAAAAZ)

[11:49] Amanda Vizedom: @MikeBennett: I should note that the capabilities I mentioned already do     (1K2AAAAAAAAAA)

exist in some development environments, especially Cyc and to a degree in some in-house, specialized     (1K2AAAAAAAAAB)

systems. But not, to my knowledge, in OWL tools generally.     (1K2AAAAAAAAAC)

[11:40] Pavithra Kenjige: @Amanda, Use Cases follow the requirement or use cases & scenarios     (1K2AAAAAAAAAD)

represent detail requirements.. which is followed by design phase. In UML, Object diagrams are done     (1K2AAAAAAAAAE)

as design stage..     (1K2AAAAAAAAAF)

[11:41] Pavithra Kenjige: Ontology development is a phase that can be mapped to design stage ..     (1K2AAAAAAAAAG)

[11:50] Peter P. Yim: == Q & A and Open Discussion, Matthew West moderating ...     (1K2AAAAAAAAAH)

[11:50] Matthew West: In the previous talk in Track C we were looking at methodologies for     (1K2AAAAAAAAAI)

integrating ontologies. The first thing that struck me was that easily the biggest priority was     (1K2AAAAAAAAAJ)

achieving consistency in the ontology, particularly in the face of large ontologies with many     (1K2AAAAAAAAAK)

contributors who are geographically dispersed. The other thing that came across was the need for an     (1K2AAAAAAAAAL)

architectural approach, providing structure to taking the right decisions in the right order.     (1K2AAAAAAAAAM)

[11:51] Matthew West: Today's session on development methods for reasoning ontologies also impressed     (1K2AAAAAAAAAN)

on me the need for an architectural approach, although the emphasis was slightly different here. The     (1K2AAAAAAAAAO)

other thing that struck me is that ontology development is often done directly at the implementation     (1K2AAAAAAAAAP)

level. You start by developing your ontology in CL or OWL, already in its implementation     (1K2AAAAAAAAAQ)

environment. This is how it was once in database development, but that was a long time ago. It would     (1K2AAAAAAAAAR)

now be considered normal to develop data models at multiple levels, a conceptual level which just     (1K2AAAAAAAAAS)

gives the outline, the logical level that is complete, including all the rules, but still     (1K2AAAAAAAAAT)

independent of any implementation environment. Finally there is a data model that takes account of     (1K2AAAAAAAAAU)

the implementation environment, and the processing needs of the application. This separation would     (1K2AAAAAAAAAV)

indicate a level of maturity in understanding the ontology development process, but will need tool     (1K2AAAAAAAAAW)

support that does not exist yet as far as I am aware.     (1K2AAAAAAAAAX)

[11:52] Doug Foxvog: Ref the capabilities Amanda referred to. The OBO ontologies have obviously not     (1K2AAAAAAAAAY)

had such tools, since, e.g., disjoint classes had had common subclasses in several posted     (1K2AAAAAAAAAZ)

[11:52] Mike Bennett: Oops! (pun intended :) )     (1K2AAAAAAAAAAB)

[11:56] Amanda Vizedom: It is good to see more tools for evaluating ontologies, including OWL if for     (1K2AAAAAAAAAAC)

no other reason than that so many people are developing ontologies in it and are in severe need of     (1K2AAAAAAAAAAD)

evaluation help. :-) It's important to note, though, that the infrastructure for easy/automated     (1K2AAAAAAAAAAE)

testing of collaborative additions to a large ontology, or addition of new ontologies to a     (1K2AAAAAAAAAAF)

repository of ontologies that are meant to be compatible, is also very much needed.     (1K2AAAAAAAAAAG)

[11:56] Mike Bennett: I should add that we have the same problems in the FIBO development - we have     (1K2AAAAAAAAAAH)

defined upper ontology partitions to distinguish e.g. independent v relative things (like the parts     (1K2AAAAAAAAAAI)

examples) but don't know when these have been misapplied until we can run external checks on it.     (1K2AAAAAAAAAAJ)

Something like what UML tools do, where illegal model efforts are flagged up as non compliant, would     (1K2AAAAAAAAAAK)

be useful in a tool.     (1K2AAAAAAAAAAL)

[11:59] Amanda Vizedom: And, along the lines of Joanne's GOEF ideas and Leo's exposition of important     (1K2AAAAAAAAAAM)

requirements questions and design decisions, tools are also very much needed to support this kind of     (1K2AAAAAAAAAAN)

analysis, and recording of the results, so that people can figure out what they need and what to     (1K2AAAAAAAAAAO)

test for in the first place.     (1K2AAAAAAAAAAP)

[12:00] Joanne Luciano: I agree w Leo - need many kinds of testing (at many levels)     (1K2AAAAAAAAAAQ)

[12:01] Joanne Luciano: Agree w/ Amanda's comments     (1K2AAAAAAAAAAR)

[12:02] Joanne Luciano: The reason I like the "component" aspect is it facilitates testing (and Leo     (1K2AAAAAAAAAAS)

just mentioned unit testing). an important "component"     (1K2AAAAAAAAAAT)

[12:01] Bobbin Teegarden: @Mike is it possible that when one starts with tree structures     (1K2AAAAAAAAAAU)

(categorization) when modeling something essentially graph shaped, that we end up in unintended     (1K2AAAAAAAAAAV)

[12:02] Doug Foxvog: @Bobbin: yes. Modeling a graph structure as a tree will almost always lead to     (1K2AAAAAAAAAAX)

[12:02] Mike Bennett: @Bobbin I think that is a real danger - unless some serious imagination is used     (1K2AAAAAAAAAAZ)

in defining abstractions "What kind of thing is this" asked iteratively until you get to a very     (1K2AAAAAAAAAAAA)

atomic meaningful concept. I don't know how you would validate / verify for that.     (1K2AAAAAAAAAAAB)

[12:04] Mike Bennett: @Bobbin plus you need to apply faceted classification, which would require some     (1K2AAAAAAAAAAAC)

additional notation. There is no reason for any given class to only have one parent (except, of     (1K2AAAAAAAAAAAD)

course, when you are designing for an application and need to think about the reasoning overhead of     (1K2AAAAAAAAAAAE)

multiple inheritance).     (1K2AAAAAAAAAAAF)

[12:08] Doug Foxvog: re Mike's reference to faceted classification: specifying the facets is     (1K2AAAAAAAAAAAG)

difficult, and ensuring those facets that should be coverings or partitions are such can be hard to     (1K2AAAAAAAAAAAH)

model in a simple ontology language. Higher-level languages such as Cyc, enable this by reifying the     (1K2AAAAAAAAAAAI)

facets as meta-classes.     (1K2AAAAAAAAAAAJ)

[12:10] Mike Bennett: @Doug thanks - I've been trying to figure out if there's a way of defining some     (1K2AAAAAAAAAAAK)

OWL annotation properties to signify facets - the obvious basis for a classification facet seems to     (1K2AAAAAAAAAAAL)

be the OWL Union Class - but need to decorate that somehow to identify what property of the parent     (1K2AAAAAAAAAAAM)

class has different values in the child classes in that facet.     (1K2AAAAAAAAAAAN)

[12:13] Doug Foxvog: @Mike: Using OWL-Full, one can create Facets & define classes as instances of     (1K2AAAAAAAAAAAO)

the facets. If several types of facets are defined.     (1K2AAAAAAAAAAAP)

[12:16] Doug Foxvog: ... If several types of facets are defined: Partition, Covering, ..., with rules     (1K2AAAAAAAAAAAR)

attached, it could force the appropriate disjointnesses and would be good for annotation, even if     (1K2AAAAAAAAAAAS)

later use of the ontology would drop such facets as mere documentation in order to allow for a     (1K2AAAAAAAAAAAT)

simpler (e.g., DL) reasoning scheme.     (1K2AAAAAAAAAAAU)

[12:21] Mike Bennett: @Doug I'll look into that - this is going to be very useful in classifying     (1K2AAAAAAAAAAAV)

financial instruments for example. Different facets are suited to different use cases, so it would     (1K2AAAAAAAAAAAW)

make sense to extract a single-inheritance taxonomy for a given use case - but different ones for     (1K2AAAAAAAAAAAX)

different use cases (e.g. investment management versus risk and compliance).     (1K2AAAAAAAAAAAY)

[12:02] Amanda Vizedom: Re: Todd's question: debugging when things don't go right - This is one     (1K2AAAAAAAAAAAZ)

reason why ontology development / management systems should have much, much more automated     (1K2AAAAAAAAAAAAA)

bookkeeping and inference traceability than many do. Again, Cyc & some other systems have this, but     (1K2AAAAAAAAAAAAB)

it is lacking in many commonly-used environments, with real consequences.     (1K2AAAAAAAAAAAAC)

[12:05] Michael Grüninger: Competency questions are great for evaluating whether or not there are     (1K2AAAAAAAAAAAAD)

enough axioms in our ontology, but there are still two outstanding issues. 1. Are these the right     (1K2AAAAAAAAAAAAE)

competency questions? 2. The competency questions themselves introduce their own ontological bias     (1K2AAAAAAAAAAAAF)

[12:05] Joanne Luciano: Agree with what Michael is saying - the competency questions aren't enough,     (1K2AAAAAAAAAAAAG)

and yes the 2nd level addresses the external requirements (Compliance standards, for example) which     (1K2AAAAAAAAAAAAH)

again is a different level than the OWL / intrinsic level.     (1K2AAAAAAAAAAAAI)

[12:06] Amanda Vizedom: @Michael +1 Questions embed some ontological commitments. That can be OK *if     (1K2AAAAAAAAAAAAJ)

it does done intentionally*, i.e., if you design your questions to test for compatibility with those     (1K2AAAAAAAAAAAAK)

commitments. But it's a real problem with many published test approaches - they embed *assumed*     (1K2AAAAAAAAAAAAL)

commitments that may not be shared in real cases in which people attempt to apply those questions.     (1K2AAAAAAAAAAAAM)

[12:06] Leo Obrst: @Matthew's comments: re: ontology development still seems to be research. Yes,     (1K2AAAAAAAAAAAAN)

that's why ontology training (a previous Ontology Summit) is very important.     (1K2AAAAAAAAAAAAO)

[12:07] Todd Schneider: To what extent are the discussions about ontology evaluations assuming a     (1K2AAAAAAAAAAAAP)

non-dynamic environment (of ontology changes)?     (1K2AAAAAAAAAAAAQ)

[12:07] Joanne Luciano: not assumed in the GOEF approach     (1K2AAAAAAAAAAAAR)

[12:08] Joanne Luciano: interesting though about autonomous systems     (1K2AAAAAAAAAAAAS)

[12:12] Amanda Vizedom: @Todd: I assume that dynamic is in fact more typical, perhaps biased by     (1K2AAAAAAAAAAAAT)

environments I have worked in. I think that much research and tool development assumes a more static     (1K2AAAAAAAAAAAAU)

model, though, and/or that individual ontologies reach a "done" stage, after which they are rarely     (1K2AAAAAAAAAAAAV)

changed and new work is on other ontologies.     (1K2AAAAAAAAAAAAW)

[12:09] Amanda Vizedom: @Matthew, @Leo -- true, but consider also the role of capturing our (ontology     (1K2AAAAAAAAAAAAX)

community) dispersed knowledge and lessons learned. Some things are better understood than others,     (1K2AAAAAAAAAAAAY)

but the understanding is unevenly distributed and redundant research continues. Or, at least, mature     (1K2AAAAAAAAAAAAZ)

hypotheses can be formed, rather than the sort proto-wheel ones that still get run up the flagpole.     (1K2AAAAAAAAAAAAAA)

[12:12] Leo Obrst: @Fabian: I agree. Please place that in the chat, so we don't lose the comment,     (1K2AAAAAAAAAAAAAB)

i.e., the 3-n things you need for evaluating ontologies.     (1K2AAAAAAAAAAAAAC)

[12:18] Fabian Neuhaus: My two questions are: (1) Assuming you need to make a recommendation to     (1K2AAAAAAAAAAAAAD)

somebody who develops an ontology. What are the three most important aspects that you think that the     (1K2AAAAAAAAAAAAAE)

person should evaluate and how? (2) Could you identify the kind of tools that would make it easier     (1K2AAAAAAAAAAAAAF)

for ontology developers to perform the recommended ontology evaluation?     (1K2AAAAAAAAAAAAAG)

[12:11] Doug Foxvog: Does anyone want to create an ontology of the ontology evaluation & development     (1K2AAAAAAAAAAAAAH)

concepts and issues that we have been discussing? With such a tool, individual ontologies could have     (1K2AAAAAAAAAAAAAI)

their properties specified using this ontology.     (1K2AAAAAAAAAAAAAJ)

[12:14] Amanda Vizedom: @doug - that is one of the hackathon & clinic proposals, though I don't know     (1K2AAAAAAAAAAAAAK)

that there will be enough interest or participation for it to pass the selection gate.     (1K2AAAAAAAAAAAAAL)

[12:15] Amanda Vizedom: Ontology of Ontology Evaluation proposal is at:     (1K2AAAAAAAAAAAAAM)

[12:17] Leo Obrst: Folks, I must go to another meeting. Thank you for your comments and     (1K2AAAAAAAAAAAAAO)

[12:18] Joanne Luciano: @LeoObrst THANK YOU -- always!     (1K2AAAAAAAAAAAAAR)

[12:19] Terry Longstreth: USE cases vs. capabilities wrt evaluation - If we assume that Use cases are     (1K2AAAAAAAAAAAAAS)

elaborated into capabilities, then evaluation based on use cases would be elaborated into more     (1K2AAAAAAAAAAAAAT)

detailed evaluations, and so on until we arrive at evaluation     (1K2AAAAAAAAAAAAAU)

[12:19] Terry Longstreth: based upon implementation     (1K2AAAAAAAAAAAAAV)

[12:18] Joanne Luciano: When I say "Function" I include in that "Capability"     (1K2AAAAAAAAAAAAAW)

[12:19] Joanne Luciano: The feedback is useful - I'll make it more explicit next time     (1K2AAAAAAAAAAAAAX)

[12:20] Joanne Luciano: words alone will often get us into trouble. -- I don't like that I use     (1K2AAAAAAAAAAAAAY)

standard for the "second" level, for example     (1K2AAAAAAAAAAAAAZ)

[12:23] Joanne Luciano: We used to use functional specifications, they worked     (1K2AAAAAAAAAAAAAAA)

[12:22] Fabian Neuhaus: @Pavithra, Matthew: maybe you could take this discussion offline?     (1K2AAAAAAAAAAAAAAB)

[12:23] Peter P. Yim: join us again, same time next week (Thu 2013.03.21), for Ontology Summit 2013     (1K2AAAAAAAAAAAAAAC)

session-10: "Software Environments for Evaluating Ontologies - II" - Co-chairs: Mike Denny & Peter P. Yim     (1K2AAAAAAAAAAAAAAD)

... please pay special attention to the start-time (for the folks in the US and Canada, unlike this     (1K2AAAAAAAAAAAAAAF)

particular session, we will back to our normal start-time next week)!     (1K2AAAAAAAAAAAAAAG)

[12:23] Peter P. Yim: great talks, thank you Joanne & Leo     (1K2AAAAAAAAAAAAAAH)

[12:24] Joanne Luciano: I want to acknowledge James Michaelis and Nicolau Depaula from CTG SUNY Albany     (1K2AAAAAAAAAAAAAAL)

for their contributions to my presentation     (1K2AAAAAAAAAAAAAAM)

[12:24] Peter P. Yim: -- session ended: 12:23pm PDT --     (1K2AAAAAAAAAAAAAAN)

-- end of in-session chat-transcript --     (1K2AAAAAAAAAAAAAAO)

Additional Resources     (1L)


For the record ...     (1L6)

How To Join (while the session is in progress)     (1M)


Attention: Please take special note on the start time of the event, as the US will be on daylight saving (summer) time on this day, while the EU is still on standard (winter) time! ... For participants from the US and Canada, for example, this session will start an hour later than usual, whereas for those from the EU, the start-time will be the same as last week's.     (1M4A)
  • Dial-in:     (1M4E)
    • Phone (US): +1 (206) 402-0100 ... (long distance cost may apply)     (1M4E1)
    • Skype: joinconference (i.e. make a skype call to the contact with skypeID="joinconference") ... (generally free-of-charge, when connecting from your computer)     (1M4E2)
      • when prompted enter Conference ID: 141184#     (1M4E2A)
      • Unfamiliar with how to do this on Skype? ...     (1M4E2B)
        • Add the contact "joinconference" to your skype contact list first. To participate in the teleconference, make a skype call to "joinconference", then open the dial pad (see platform-specific instructions below) and enter the Conference ID: 141184# when prompted.     (1M4E2B1)
      • Can't find Skype Dial pad? ...     (1M4E2C)
        • for Windows Skype users: Can't find Skype Dial pad? ... it's under the "Call" dropdown menu as "Show Dial pad"     (1M4E2C1)
        • for Linux Skype users: please note that the dial-pad is only available on v4.1 (or later; or on the earlier Skype versions 2.x,) if the dialpad button is not shown in the call window you need to press the "d" hotkey to enable it. ... (ref.)     (1M4E2C2)
  • Shared-screen support (VNC session), if applicable, will be started 5 minutes before the call at: http://vnc2.cim3.net:5800/     (1M4F)
    • view-only password: "ontolog"     (1M4F1)
    • if you plan to be logging into this shared-screen option (which the speaker may be navigating), and you are not familiar with the process, please try to call in 5 minutes before the start of the session so that we can work out the connection logistics. Help on this will generally not be available once the presentation starts.     (1M4F2)
    • people behind corporate firewalls may have difficulty accessing this. If that is the case, please download the slides above (where applicable) and running them locally. The speaker(s) will prompt you to advance the slides during the talk.     (1M4F3)
    • instructions: once you got access to the page, click on the "settings" button, and identify yourself (by modifying the Name field from "anonymous" to your real name, like "JaneDoe").     (1M4G1)
    • You can indicate that you want to ask a question verbally by clicking on the "hand" button, and wait for the moderator to call on you; or, type and send your question into the chat window at the bottom of the screen.     (1M4G2)
    • thanks to the soaphub.org folks, one can now use a jabber/xmpp client (e.g. gtalk) to join this chatroom. Just add the room as a buddy - (in our case here) summit_20130314@soaphub.org ... Handy for mobile devices!     (1M4G3)
  • Discussions and Q & A:     (1M4H)
    • Nominally, when a presentation is in progress, the moderator will mute everyone, except for the speaker.     (1M4H1)
    • To un-mute, press "*7" ... To mute, press "*6" (please mute your phone, especially if you are in a noisy surrounding, or if you are introducing noise, echoes, etc. into the conference line.)     (1M4H2)
    • we will usually save all questions and discussions till after all presentations are through. You are encouraged to jot down questions onto the chat-area in the mean time (that way, they get documented; and you might even get some answers in the interim, through the chat.)     (1M4H3)
    • During the Q&A / discussion segment (when everyone is muted), If you want to speak or have questions or remarks to make, please raise your hand (virtually) by clicking on the "hand button" (lower right) on the chat session page. You may speak when acknowledged by the session moderator (again, press "*7" on your phone to un-mute). Test your voice and introduce yourself first before proceeding with your remarks, please. (Please remember to click on the "hand button" again (to lower your hand) and press "*6" on your phone to mute yourself after you are done speaking.)     (1M4H4)
  • RSVP to peter.yim@cim3.com with your affiliation appreciated, ... or simply just by adding yourself to the "Expected Attendee" list below (if you are a member of the community already.)     (1M4J)
  • Please note that this session may be recorded, and if so, the audio archive is expected to be made available as open content, along with the proceedings of the call to our community membership and the public at-large under our prevailing open IPR policy.     (1M4L)

Attendees     (1N)



This page has been migrated from the OntologWiki - Click here for original page     (1N4)