From OntologPSMW

Jump to: navigation, search
[ ]
Session Synthesis Session 2
Duration 1.5 hour90 minute
5,400 second
0.0625 day
Date/Time Apr 26 2017 18:30 GMT
9:30am PDT/12:30pm EDT
5:30pm BST/6:30pm CEST
Convener KenBaclawski

Ontology Summit 2017 Synthesis Session 2     (2)

Meeting ID: 768423137     (4)

Please use the chatroom above. Do not use the video teleconference chat, which is only for communicating with the moderator.     (6)

When you use the Video Conference URL above, you will be given the choice of using the computer audio or using your own telephone. Some attendees had difficulties when using the computer audio choice. If this happens to you, please leave the meeting and reenter it using the telephone choice. You will be given a telephone number to call along with an access code.     (7)


Abstract     (8)

The primary aim of this week's session is to synthesize the lessons learned so far on the three tracks that are under way. Each track has met twice, and so we will have gained insights from a combination of invited speakers, chat log comments and blog page discussions. The outcome of today's session will form the basis of this year's Ontology Summit Communiqué.     (8A)

A secondary goal is to formulate challenge problems both broadly for the Communiqué and more narrowly as challenges for the Symposium.     (8B)

Agenda     (9)

Attendees     (10)

Proceedings     (11)

[12:15] KenBaclawski: Video Teleconference:     (11A)

[12:31] Donna Fritzsche: Hi all!     (11B)

[12:38] MikeBennett: What is that thing above? It turns up every week now     (11C)

[12:42] ravisharma: I can not dial in as meeting audio code is not acceptable 510720#     (11D)

[12:42] MikeBennett: @Ravi meeting code is 768423137     (11E)

[12:43] ravisharma: thanks     (11F)

[12:50] Ken Baclawski: @MikeBennett: There is a blinking red dot in the upper right corner of the screen to show that the session is being recorded. Is that what you mean?     (11G)

[12:51] MikeBennett: No I was wondering about the anonymous111111111 bomb     (11H)

[12:53] Ken Baclawski: @MikeBennett: It seems to be an "undocumented feature" of soaphub...     (11I)

[12:53] ToddSchneider: Has the notion of 'persisting knowledge' [gained from machine learning or other paradigms] and how ontologies meet such a need been brought up?     (11J)

[13:00] BobbinTeegarden: Within an ontological scaffolding, are there subsets of sense, contexts like rooms?     (11K)

[13:00] Donna Fritzsche: good question Bobbin     (11L)

[13:00] ToddSchneider: The notion of 'context' is critical to developing and using/reusing ontologies, but is still an open research problem.     (11M)

[13:02] ToddSchneider: By 'ontological scaffolding' is this synonymous with Ontology Architecture?     (11N)

[13:02] Donna Fritzsche: Clarion system - seems like a great tie in to the reasoning track     (11O)

[13:02] Donna Fritzsche: Gary - can you provide a reference?     (11P)

[13:04] Donna Fritzsche: Gary mentions a key point - Training data is difficult to obtain - for label acquisition.     (11Q)

[13:04] ToddSchneider: Gary, could you update your slides by replacing 'DL' with 'deep learning' where applicable (otherwise a reader may interpret 'DL' as description logic)?     (11R)

[13:04] Ken Baclawski: @ToddSchneider: Situations are one approach to context that has been successful in some domains. While situations are not a general solution to context, they could be the basis for such a solution.     (11S)

[13:04] Donna Fritzsche: @todd - do you have a reference for Situations     (11T)

[13:05] ravisharma: Gary - could these algorithms be understood in terms of intermediate neural patterns and neural signal flow processes before relating them to human understandable items?     (11U)

[13:05] ToddSchneider: Donna, Ken has written about situations and there's also a book/papers by Barwise and Devlin.     (11V)

[13:06] Donna Fritzsche: thanks! interesting topic Ken - I was not aware of this work - my apologies!     (11W)

[13:07] Donna Fritzsche: Ontological Scaffolding Brainstorm exercises could be an interesting topic for our in-person workshops.     (11X)

[13:07] ravisharma1: Yes Donna     (11Y)

[13:08] ToddSchneider: What does 'Ontological Scaffolding' refer to (i.e., how should I interpret this phrase)?     (11Z)

[13:09] ravisharma1: a poster that describes ontologies in simplest level     (11AA)

[13:09] Donna Fritzsche: @Gary and Bobbin - could either of you provide us a good definition of Ontological Scaffolding - or is the definition still a brainstorm exercise. Gary talked about briefly, just now.     (11AB)

[13:10] gary berg-cross: @Todd 13:02 No By 'ontological scaffolding' is not synonymous with Ontology Architecture but it would have an architecture.     (11AC)

[13:11] KenBaclawski: @Donna: My talk in Track B has some references. See also     (11AD)

[13:11] ToddSchneider: Gary, so how should 'Ontological Scaffolding' be understood?     (11AE)

[13:12] gary berg-cross: @Donna there is a chapter on CLARION (Ron Sun from,     (11AF)

[13:12] ToddSchneider: Gary, see my request at 13:02.     (11AH)

[13:13] gary berg-cross: @Todd I know about the DL confusion but people in ANNs use it regularly.     (11AI)

[13:13] ToddSchneider: So, it 'Ontological Scaffolding' some sort of systems architecture?     (11AJ)

[13:14] ToddSchneider: Gary, but our context is more that ANN (whatever that refers to). Some of us are acronym challenged:)     (11AK)

[13:15] Donna Fritzsche: @gary, all - would ontological scaffolding be more about the lifecycle/progression in the growth of the ontology. starting for instance with a seed set (NELL) and growing thru different stages and by different methods over time?     (11AL)

[13:15] ChristiKapp: @Donna is 'Ontological Scaffolding' more like a process framework that guides ontology architecture development? (referring by analogy to sensorimotor stages and cognitive scaffolding)     (11AM)

[13:15] gary berg-cross: @ Ravi "could these algorithms be understood in terms of intermediate neural patterns and neural signal flow processes?" If I understand this relates a neural pattern to some other type of "thing" whether concept or algorithm. It helps to relate things but they are different.     (11AN)

[13:15] Donna Fritzsche: @christi Jinks!     (11AO)

[13:16] Donna Fritzsche: @christi - yes, I agree with you     (11AP)

[13:18] gary berg-cross: @Todd How should'Ontological Scaffolding'be understood? Think of it as a starter set which in normal operations can develop into more knowledge and reasoning.     (11AQ)

[13:19] ToddSchneider: Gary, 'normal operations'? Of, or on, what? This sounds like a process.     (11AR)

[13:20] ravisharma1: Mike - they are also using IBM Watson to help discover phishing and fraud type activities in gov.     (11AS)

[13:22] ToddSchneider: As a really practical application the technologies described during this year's summit would be application to U.S. federal laws and regulations to help be able to understand their imapacts.     (11AT)

[13:23] Donna Fritzsche: on a somewhat related note - identifying "fake news" would be an interesting challenge - and explaining why it is fake.     (11AU)

[13:24] BobbinTeegarden: How about inspecting tweets to predict what Trump is going to say next?     (11AV)

[13:24] Donna Fritzsche: I have thought of that one also Bobbin. I think it might be out there - a Trump tweet bot     (11AW)

[13:25] gary berg-cross: On this Q of cognitive scaffolding and essential cores people as of humans anyway what is the maximum amount of working memory required to perform a task? What is the maximum amount of mental computation required what to do next? How much task-relevant information is encoded in the environment, and how easily is it used. To date, there is no general theory but groups like NELL have an environment where they may test out some ideas.     (11AX)

[13:27] Ram D. Sriram: @Donna, Robin: Check out     (11AY)

[13:27] gary berg-cross: @Todd Normal ops will vary from AIs, some of which are embeded like robots and humans but the general idea is that they agent-environment state machines, i.e. systems whose cognitive states are distributed between internal mechanisms, external settings (beliefs, goals intentions etc.), and the resulting interaction between internal and external dynamics.     (11AZ)

[13:30] ToddSchneider: Gary, still sounds like an ontology architecture: A lattice of ontologies that increasingly constrain interpretations for their application domains.     (11AAA)

[13:32] BobbinTeegarden: Context is a (potentially complex) domain constraint in the large?     (11AAB)

[13:32] ToddSchneider: Bobbin, definitely.     (11AAC)

[13:33] gary berg-cross: @Mike On this topic of verbal deception and hiding meaning by creative use of terms, one approach we investigated on this some years ago was to use cohesive measures of how thing fit together to give a whole sense. It is a bit of an abductive reasoning exercise.     (11AAD)

[13:34] ToddSchneider: One of problems with attempting to formalize 'context' is its scope and dynamism.     (11AAE)

[13:35] BobbinTeegarden: So how do you model context? It's a snarl of and-sets within a wider structured ontology?     (11AAF)

[13:37] BobbinTeegarden: @Todd ... and how do you model the dynamism in 2D (or 3D)?     (11AAG)

[13:38] ToddSchneider: Bobbin, as I mentioned what a context is, is an open research problem.     (11AAH)

[13:38] gary berg-cross: Context is a well dicussed topic on the Ontolog Forum and more broadly. A starting point for understanding alternative formulations might be     (11AAI)

[13:39] ToddSchneider: Bobbin, how you model or represent 'something' depends on the intended uses of the model/representation.     (11AAJ)

[13:39] ChristiKapp: @Donna - I wonder if "fake news" is misnamed. If each group of humans (e.g. left or right) possesses knowledge and truth with contextual boundaries that limit their thoughts and reasoning - then something that does not logically fit within their context is perceived as "fake". In order to detect it, knowledge of who created the message is important - because there are networks of news-creators sitting outside of contextual boundaries creating "news with a particular intention" of reinforcing knowledge within a left or right leaning contextual boundary.     (11AAK)

[13:40] ravisharma: Ram - we again lost the right slides, it is not synchronized with your talking points.     (11AAL)

[13:40] Donna Fritzsche: @christi - great points, but there are also those that literally generated "fake news" with the intent to deceive/miscommunicate     (11AAM)

[13:40] ToddSchneider: Bobbin, 'intended uses' of 'something' implicitly embed aspect of the usage context (e.g., assumptions about the usage environment).     (11AAN)

[13:41] BobbinTeegarden: @Todd Yes!     (11AAO)

[13:42] MikeBennett: @Todd @Bobbin exactly - context is not so much a feature of meaning, it's a feature of "the meaning of" some word, term etc.; so in this context, what concept does the symbol "bear" refer to.     (11AAP)

[13:42] ToddSchneider: Bobbin, note that what I've expressed in more of an engineering perspective, even though metaphysics may be more entertaining.     (11AAQ)

[13:42] Donna Fritzsche: @christi - It would be interesting to look a news stories and try to identify the world views of who generated them. There could be a decent training set/unknown set pulling from known news sources with known biases,etc     (11AAR)

[13:42] BobbinTeegarden: @Gary Thank you for that URL Pat Hayes on context. Fun.     (11AAS)

[13:43] ravisharma: Ram - it is fine now.     (11AAT)

[13:44] David Newman: Mike, good explanation on vectors re: semantic distance.     (11AAU)

[13:44] ToddSchneider: What are the differences between a (common understanding of a) 'knowledge graph' and an 'ontology'?     (11AAV)

[13:44] ChristiKapp: @donna - yes I agree - they do that in order to deceive those inside a particular context into thinking something specific about those whose reasoning process is performed in the other context. the fake news process makes use of the human "habit loop" also     (11AAW)

[13:45] MikeBennett: Thanks @DavidNewman I didn't feel I did justice to your excellent explanation the other week but I'm glad I caught the essence of it.     (11AAX)

[13:45] ToddSchneider: Mike, we really shouldn't be using the term 'meaning'. Instead the term 'interpretation' better represents 'our' intention.     (11AAY)

[13:46] MikeBennett: @Todd agreed. When people say things like "meaning is context" I don't think this is a statement about concepts, but about interpretation, as you say.     (11AAZ)

[13:46] JackRing: I am enthused by the synergy arising in this dialog. Looks to me like an way of understanding a computer program and particularly arbitrarily large, heterogeneous suites of interacting computer programs is on the horizon. Big Data is interesting and valuable but Big Code is the far larger opportunity --- and the solution to cybersecurity. I realize that this may not fit in this year's focus locus but perhaps one or more of you are interested in exploring this.     (11AAAA)

[13:47] BobbinTeegarden: @JackR Ah, thank you Jack -- we have Big Data, but where is Big Process?     (11AAAB)

[13:49] BobbinTeegarden: Is the context we have been talking about somehow related to the 'process' i.e. ontology usage? Is that a sub-ontology extract?     (11AAAC)

[13:50] Donna Fritzsche: @Christ - yes - interesting     (11AAAD)

[13:51] ToddSchneider: Bobbin, in some cases yes.     (11AAAE)

[13:51] MikeBennett: @Bobbin not if the ontology is done right. Use of words is contextual; ontology should be the tool to get past that by framing the concepts.     (11AAAF)

[13:53] JackRing: Do not confuse ontology usage with entries in an ontology that denote logic, arithmetic and semantic relations in a computer program. Think of an ontology of a given set of code, then of another then how to determine the coherence of the respective ontologies.     (11AAAH)

[13:53] MikeBennett: Isn't Knowledge Graph a set of assertions at the level of individuals, ontology is the concepts (TBox); some people use the word to mean any OWL file even if it is ABox level but I find that nhelpful     (11AAAI)

[13:53] ChristiKapp: @Donna I wonder if "dark learning" is just "learning performed in a context that an outside observer doesn't understand because of mis-aligned context? So people as a group are seeing "dark learnings" of other groups and not understanding each other's perspectives anymore. Does the dark learning apply to groups of people as well as AI?     (11AAAJ)

[13:53] Donna Fritzsche: @Bobbin - Big Process - might be an interesting topic for our workshops - we talked about using idealized "lego" like abstractions of the different processes/tools employed during the talk and tackling new problems by combining them. In doing so, we also could identify important touchpoints/ areas where cleaner apis need to be made     (11AAAK)

[13:53] JackRing: Meanwhile, closer to this year's goals, I am disappointed at the lack of attention to pathologies and pathogens in the current triad.     (11AAAL)

[13:54] ToddSchneider: Jack, 'pathologies and pathogens' in what context?     (11AAAM)

[13:54] TerryLongstreth: @Jack - +1 Pathogenic outcomes of using or building ontologies     (11AAAN)

[13:55] BobbinTeegarden: @Donna legos as patterns, or legos forming 'contextual' patterns, or both? Is it fractal, in the end?     (11AAAO)

[13:56] Ram D. Sriram: I am logging off.     (11AAAP)

[13:57] Donna Fritzsche: @bobbin - legos as tools/pieces of functionality - Big Process and then we could have legos as context patterns - providing the backdrop in which these processes operation. (yes, I would say it is/can be fractal in the end!)     (11AAAQ)

[13:57] Donna Fritzsche: Data: Context: Process Models     (11AAAR)

[13:58] MikeBennett: Context can also be a synonym for the notion of "Mediating Thing" in Sowa's formulation i.e. that in which some set of things are brought together in specific roles or functions. So Sales is the context for Customer - it would not exist as a concept without that context.     (11AAAS)

[14:00] TerryLongstreth: Better wording of my 13:54: Pathogenic consequences of using or building ontologies     (11AAAT)

[14:01] Donna Fritzsche: Thanks!     (11AAAU)

[14:01] ravisharma: thanks bye for now     (11AAAW)

[14:02] ToddSchneider: Meeting ended at 14:00 EST.     (11AAAX)

[14:02] BobbinTeegarden: @Gary thank you again, and oh no... back to mereology. Uroboros!     (11AAAY)

[14:02] gary berg-cross: Hayes thesis proposed some time ago (Hayes 1997): that there is no useful single idea of "context", and in order to construct nontrivial theories of contextual truth, it is necessary to distinguish different conceptions of context and analyze their varying and different properties. In this spirit, then, the formal results are offered as a contribution towards a useful theory of spatiotemporal contextualization, with an accompanying suggestion that it might be useful to try to characterize the basic assumptions of other notions of context and contextualization of truth in the same formal, axiomatic style.     (11AAAZ)

[14:05] BobbinTeegarden: @gary  !     (11AAAAA)

[14:10] ravisharma: Gary - is that what would limit context to the present tense only or that it would also allow stories embedded among stories?     (11AAAAB)

[14:11] ravisharma: i guess Gary has left     (11AAAAC)

Resources     (12)

Previous Meetings     (13)