From OntologPSMW

Jump to: navigation, search
[ ]
    (1)
Session Ernest Davis
Duration 1 hour
Date/Time 06 May 2020 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CEST
Convener Ram D. Sriram
Track Why

Contents

Knowledge graphs, closely related to ontologies and semantic networks, have emerged in the last few years to be an important semantic technology and research area. As structured representations of semantic knowledge that are stored in a graph, KGs are lightweight versions of semantic networks that scale to massive datasets such as the entire World Wide Web. Industry has devoted a great deal of effort to the development of knowledge graphs, and they are now critical to the functions of intelligent virtual assistants such as Siri and Alexa. Some of the research communities where KGs are relevant are Ontologies, Big Data, Linked Data, Open Knowledge Network, Artificial Intelligence, Deep Learning, and many others.     (2A)

Agenda     (2B)

Conference Call Information     (2C)

Attendees     (2D)

Discussion     (2E)

[12:11] Ravi Sharma: Ernie - why is there a problem in temporal reasoning? Is it calendars is it not being decimal way of counting time or is it imprecise definition of how time is placed in a NL question?     (2E1)

[12:15] Ravi Sharma: Ernie - even though you defined temporal standards, is it lack of proper use or lack of a preferred standard of time to be followed?     (2E2)

[12:17] Ravi Sharma: Ernie - is context playing a role or concatenation, it seems all over?     (2E3)

[12:19] Ravi Sharma: Ernie - the problem of spatial resolution is always there also!     (2E4)

[12:23] Ravi Sharma: Ernie - why do the AI programs not normalize various temporal expressions shown by you into a standard of time such as UT GMT and go with reasoning thereafter?     (2E5)

[12:27] Leia Dickerson: Couldn't one address issues of outdated information, in part, by looking at the metatdata for pages that information is posted on? The example you showed had a date stamp for when the information was accurate for.     (2E6)

[12:29] David Eddy: @Leia... very dangerous to rely on date stamps on "documents"... particularly anything touched by Microsoft products... by viewing a document, that can change the modification date.     (2E7)

[12:30] Leia Dickerson: @There's a difference from a "created date" and an edited date.     (2E8)

[12:30] Leia Dickerson: @David     (2E9)

[12:30] Leia Dickerson: This is what Ravi's question is getting at it seems.     (2E10)

[12:30] Ravi Sharma: Ernie -is there a freq analysis of which standard of time is useful?     (2E11)

[12:31] Todd Schneider: From Nancy Wiegand to Everyone: (12:29 PM) How are you going to represent time, say in a triple format or RDB format or what?     (2E12)

[12:31] David Eddy: @Leia... I know.. I've seen MSWord change the edit date, simply viewing it. Behavior may have changed... who knows.     (2E13)

[12:32] Leia Dickerson: It definitely depends on what type of material you are looking at.     (2E14)

[12:33] Leia Dickerson: HTML pages are different from Word documents, which are different from PDFs, (on the open web) which are different from materials from bibliographic databases which have controlled metadata.     (2E15)

[12:33] JoshL: OWL-time     (2E16)

[12:33] JoshL: W3c     (2E17)

[12:34] Mark Underwood @knowlengr: I think you are looking at the wrong Microsoft metadata; their cybersecurity temporal tagging is satisfactory; one doesn't look inside the document for that     (2E18)

[12:34] Mark Underwood @knowlengr: See the windows event log, security group, verbose     (2E19)

[12:35] Ravi Sharma: Ken - Leia has lot of comments Qs     (2E20)

[12:35] David Eddy: @MarkU.... >> cybersecurity temporal tagging is satisfactory << do "civilians" using run-of-the-mill O365 have access to cybersecurity level functionality?     (2E21)

[12:36] TerryLongstreth: In the space domain we have to worry about relativistic effects (time compression and dilation). So time "grounding" requires capture of the position/velocity vector (against a reference spatial point or body) for any time value stored. Has this issue been examined w/r/t elementary data types referred to in today's discussion?     (2E22)

[12:37] Mark Underwood @knowlengr: Time is complicated, yes, @Ken, in part because events are complicated (cf. complex event processing)     (2E23)

[12:37] Mark Underwood @knowlengr: One could argue that "domain" is always in play     (2E24)

[12:38] JoshL: Event time often has at least three different values - the time of the event, the time at which that time is asserted or predicted, and the time interval over which the assertion is valid.     (2E25)

[12:39] Janet Singer: @JoshL good point     (2E26)

[12:40] Ravi Sharma: Todd thanks - I copies from Nancy: How are you going to represent time, say in a triple format or RDB format or what?     (2E27)

[12:40] Mark Underwood @knowlengr: @Josh Depending on the domain, I don't think that simplification is helpful     (2E28)

[12:40] Janet Singer: @Mark ?     (2E29)

[12:41] Mark Underwood @knowlengr: @Janet most events of human interest are composites of other events... The "time" of the event has to be spelled out by the domain expert, and is unlikely to be expressed in this simple form     (2E30)

[12:43] Mark Underwood @knowlengr: @Janet Topic of previous summits     (2E31)

[12:45] Mark Underwood @knowlengr: @david Cybersecurity is not a lay person specialization; the obscurity of the event log viewer reflects indifference to this     (2E32)

[12:45] Ravi Sharma: for many applications such as repeating cycles of satellite orbiting images, IoT data etc space is in control i.e. for closed systems, but in open conversation, ambiguities in dealing with space can be challenging     (2E33)

[12:46] Gary Berg-Cross: What do your think are the common-sense knowledge problems with the concept of "feature" as in "spatial feature" or "geographic feature" or "geological feature"?     (2E34)

[12:46] Ravi Sharma: NASA Deals it through metadata     (2E35)

[12:47] Ravi Sharma: I meant it for Ernie (speaker)     (2E36)

[12:49] Janet Singer: @Mark, Agree that domain is always in play, but isn't a good starting assumption about the domain of assertions on the internet that they all have those three relevant time aspects?     (2E37)

[12:50] Ravi Sharma: Ernie - 3D models for Manufacturing the item such as Pyramid example can fully handle making, maintaining, materials and supply chain and sequencing. No problems in engineering space.     (2E38)

[12:50] Ravi Sharma: space     (2E39)

[12:51] David Eddy: I'm trying to remember name of a Y2K friend who had a tool able to identify something like 130+ different date formats.     (2E40)

[12:52] Ravi Sharma: string bag is solvable by fluid dynamics solvers such as Fluent, MatLab     (2E41)

[12:53] Evan Wallace: 2nd the Gary's question about features with respect general geometry models. This is especially interesting if you include holes/spaces as geometric features.     (2E42)

[12:53] Mark Underwood @knowlengr: lol I like the idea of a hallucinating AI     (2E43)

[12:54] Mark Underwood @knowlengr: @Janet yes, a sensible starting point but only if the domain ontology is already pretty mature     (2E44)

[12:55] Ravi Sharma: Ernie - there are lot of work done in image processing and intelligence that can find a needle in haystack?     (2E45)

[12:56] Mark Underwood @knowlengr: @Janet worth a deeper dive sometime [sic]     (2E46)

[12:56] Todd Schneider: The notion 'Feature' and the term 'Feature' should be avoided in ontologies.     (2E47)

[12:56] Janet Singer: JoshLs point is three standardizable ways time is relevant conceptually (time of assertion, time asserted, time of validity), while Davids point is of token varieties capturing those concepts     (2E48)

[12:58] ABollans: @Ravi this assumes you have a nice 3D model     (2E49)

[12:58] Ravi Sharma: there are algorithms that might identify illogical or incomplete info such as non closed polygons     (2E50)

[12:58] Mark Underwood @knowlengr: Sorry, off to next meeting, practice keeping a safe space/time from others     (2E51)

[12:59] Gary Berg-Cross: I hope that Michael G has a Q to further discussion.     (2E52)

[12:59] JoshL: Knowledge representation and processing for mixed reality visualization concerns itself a great deal with relative spatial and temporal positioning of multiple actors. Where do you see frames of reference coming into the representations you have discussed?     (2E53)

[12:59] Ravi Sharma: for astronomy different scales of time and space apply.     (2E54)

[13:09] Nancy Wiegand: I'm curious as to how time or other kinds of relevant information is handled in a triple. For example, a river is called a 'kill' in the eastern U.S. but not elsewhere. How is this represented in a triple?     (2E55)

[13:12] Ravi Sharma: Ernie - thanks for your talk on what knowledge graphs with some AI can improve the situations relating to time and space     (2E56)

[13:13] Todd Schneider: Meeting ends @13:13 EDT     (2E57)

Resources     (2F)

Previous Meetings     (2G)


Next Meetings     (2H)