From OntologPSMW

Jump to: navigation, search
[ ]
Session Synthesis I
Duration 1 hour
Date/Time 25 March 2020 16:00 GMT
9:00am PDT/12:00pm EDT
4:00pm GMT/5:00pm CET
Convener Ken Baclawski


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)

The session will be an open discussion. We encourage everyone to share their opinions and points of view. Given the large amount of material that will have to be debated and synthesized, next week will also be a Synthesis session. Video Recording YouTube Video     (2B1)

Conference Call Information     (2C)

Attendees     (2D)

Discussion     (2E)

[12:05] David Eddy: I am not certain if the issue of "unnatural language" is considered in or out of KG discussion...     (2E1)

[12:06] David Eddy: accessibility is new... KG now available at desktop level     (2E2)

[12:09] David Eddy: "Ontologies" are about as well understood as "metadata"     (2E3)

[12:10] John Sowa: There are many, many different words.     (2E4)

[12:10] John Sowa: We can use 'logic' and 'ontology' as the basic terms.     (2E5)

[12:11] John Sowa: Then relate those terms to all the others that people use.     (2E6)

[12:12] Ravi Sharma: We still have not much convergence on what are KGs, essential elements minimum expectation and at least common understanding by communities what they mean by KGs?     (2E7)

[12:12] ToddSchneider: Semantic Coherence allows for Interoperability.     (2E8)

[12:13] Ravi Sharma: John likes DOL standard.     (2E9)

[12:13] David Eddy: One of my favorite acquisitions... it took some 200 years from Isaac Newton doing alchemy to appearance of Mendeleev's Periodic Table     (2E10)

[12:14] Ram D> Sriram: @Ravi: I think one way to look at it is Where we are now, where do we want to go, and how do we get there. Standards may play an important in getting us there.     (2E11)

[12:15] Ravi Sharma: CL and FOL relate to KGs according to John     (2E12)

[12:16] Ravi Sharma: John - Models, we can relate to them by math and logic.     (2E13)

[12:17] Ravi Sharma: John begins with Conceptual models and so on     (2E14)

[12:17] Ravi Sharma: John says semantic model and logic same?     (2E15)

[12:18] Mark Underwood @knowlengr: It is a sidebar track, but the meetings I've been able to attend have been weak in leveraging the graph tools that are powering the current interest. E.g., yesterday's webinar     (2E16)

[12:18] David Eddy: I would argue that in-use AS IS "models" have not been built/maintained clear semantics     (2E17)

[12:18] Mark Underwood @knowlengr: i.e., TigerGraph, Neo4J etc     (2E18)

[12:19] David Eddy: an intermediate step to formally acknowledge / define in-use terminology, which then can be mapped to more formal representations     (2E19)

[12:19] Ravi Sharma: successive steps at refinement.     (2E20)

[12:21] John Sowa: What we need is a glossary. logic and ontology are the basis for precise definitions.     (2E21)

[12:22] David Eddy: @JFS... MANY local glossaries, sitting next to each application.     (2E22)

[12:22] John Sowa: But the glossary should include all the popular terms that are being kicked around in every approach.     (2E23)

[12:23] John Sowa: David, we need an Ontolog glossary. It should include the defining words of everybody's application glossaries.     (2E24)

[12:23] David Eddy: @JFS... "ALL" gets to be a very big number     (2E25)

[12:24] David Eddy: @JFS... ok... you're focusing on an ontology glossary     (2E26)

[12:27] Paul Tyson: Reality check: your boss says to make a knowledge graph that will augment your organization's effectiveness, and give me a one-page plan for it tomorrow. What do you do?     (2E27)

[12:27] janet singer: Ontology should not be the foundation     (2E28)

[12:27] ToddSchneider: A very important tool in ontology are annotations: provide out-of-band (aka meta) information that describe the intention of the representation (for each notion in the ontology).     (2E29)

[12:29] ToddSchneider: Is 'knowledge' actionable information or does the use of 'knowledge' allow for 'actionable information'?     (2E30)

[12:30] Gary Berg-Cross: Some of the things to address in the Communique (based in part by what we understand from the Uses cases and an ontological/KR point of view: 1: Capturing Context, 2: Domain-Specific Knowledge Extraction to populate KGs, 3: Data/Knowledge Alignment & harmonization (includes with ontologies), 4: Near Real-Time KG for BIG/Fast Data (like KGs for disasters), 5: providing Quality and Validity of KGs, 6: Building in Adaptive process for KGs in an (Open?) Knowledge Network, 7. KGs maybe as part of an OKN that employs ML and in turn supports ML. (based on ideas in     (2E31)

[12:31] ToddSchneider: Is it the case that 'knowledge' is a priori information and 'actionable information' a posteriori information?     (2E32)

[12:31] John Sowa: Janet, I'll drop the world 'foundation'. It's more confusing than helpful.     (2E33)

[12:32] John Sowa: Paul, when your boss gives you an assignment, you use whatever tools your business has adopted.     (2E34)

[12:32] David Eddy: @Ravi - "active" vs "passive" is an old, old, old challenge in the long forgotten world of central data dictionary from the 1960s.     (2E35)

[12:34] John Sowa: But somebody in your business has to learn and develop the methods for relating the local tools to the tools for the DOL standard.     (2E36)

[12:35] ToddSchneider: Use of a graph representation allows for more 'complete' representation of a domain.     (2E37)

[12:35] David Eddy: @Ravi... but what are your assumptions about said source on left? Clear, well defined/documented terms & definitions?     (2E38)

[12:37] John Sowa: Todd, the word 'complete' is ambiguous. In one sense, the Common Logic version is the most general, since it can relate any or all of the other versions.     (2E39)

[12:37] John Sowa: But in another sense, the ordinary language explanation is the most complete because people can relate it to everything they know and do.     (2E40)

[12:38] John Sowa: That is why we need a glossary to relate all the terms to one another.     (2E41)

[12:38] David Eddy: @Janet... do we include or exclude the systems that make the data... or we only talking about data.     (2E42)

[12:39] ToddSchneider: John, yes 'complete' is ambiguous. The intent was to suggest that use of a graph for representing information provides both a better solution (from an operational perspective) and representing more of a domain (e.g. more relations).     (2E43)

[12:40] BobbinTeegarden: DEC also had an Active Data Dictionary in the 80s     (2E44)

[12:41] Ravi Sharma: David - it still exists, in terms of data models, vocabularies and governance rules e.g. HIPAA     (2E45)

[12:42] janet singer: @David: If we take the basic cycle as Data > Inferencing > Actionable Info > [action] > Data we can refine and elaborate that as appropriate     (2E46)

[12:42] David Eddy: @Bobbin... yes... CDD+     (2E47)

[12:42] Leia Dickerson: Librarianship has methods for this:     (2E48)

[12:43] Paul Tyson: You wanna glossary? Fire up emacs and bang out an RDFS file. Next problem?     (2E49)

[12:43] David Eddy: @Paul... not A glossary... MANY glossaries.     (2E50)

[12:44] David Eddy: with mapping between them when needed     (2E51)

[12:44] Paul Tyson: Wash, rinse, repeat.     (2E52)

[12:45] Doug Holmes: @ Janet - not all systems of interest, that make use of internally represented human knowledge, have no or very little need to communicate with humans. In this case, Actionable Info does not seem to be a requirement, but intelligent behavior is     (2E53)

[12:46] janet singer: As I said it applies to all data-and-model-driven actors     (2E54)

[12:46] David Eddy: @JFS... do we therefore include unnatural terms such as IBM... or are we constrained to International Business Machines?     (2E55)

[12:46] Ravi Sharma: john- KGs have following elements - logic,models, information sources, NL as well non-NL e.g. software and glossary which john seems to define something similar to programmers coding rules so we all stay at same objective and solutions that work as system as a whole?     (2E56)

[12:47] BobbinTeegarden: At OMG yesterday had a presentation on NEIM presenting a 'model' with reuseable/revisable 'Lego blocks' wrapped in 'tinker toy' wrappers -- choose your context and glue together     (2E57)

[12:47] janet singer: @David I would say the machines use actionable info as well     (2E58)

[12:47] Mark Underwood @knowlengr: @Gary add cybersec attack and risk graphs to use case catalog     (2E59)

[12:48] Gary Berg-Cross: The idea of work group committee activities. Groups could report back at the next synthesis session.     (2E60)

[12:49] David Eddy: "glossary" when easily accessible is excellent mechanism to point out my understanding of a term is not same as your...     (2E61)

[12:49] Paul Tyson: @Ravi my KGs only have triples.     (2E62)

[12:49] Mark Underwood @knowlengr: @Gary It's sorta adjacent, but much-invested-in -- social network graphs (Facebook, LinkedIn, phishing assessments)     (2E63)

[12:49] John Sowa: Gary, the Ontolog glossary would be the map between all the ways of talking, writing, and implementing.     (2E64)

[12:49] BobbinTeegarden: NEIM as an ontology. Wrappers give recursion.     (2E65)

[12:50] Mark Underwood @knowlengr: @Bobbin NEIM still alive?     (2E66)

[12:50] Leia Dickerson: What time commitment do you imagine for this first phrase for reporting out next week?     (2E67)

[12:50] John Sowa: I'm currently working on a dialect of Common Logic that I call CLIP.     (2E68)

[12:50] David Eddy: Would I be a work-group of one on topic of "un-natural language"?     (2E69)

[12:50] BobbinTeegarden: @Mark NEIM V3 having a new breath of life, we'll see...     (2E70)

[12:51] Ravi Sharma: i can visualize repositories openly accessible but actionable if required both working to provide knowledge as required but would that need kg     (2E71)

[12:51] Mark Underwood @knowlengr: @john <g> remember CLIPS from the 80's?     (2E72)

[12:51] John Sowa: I'm also writing the slides for relating CLIP to knowledge graphs.     (2E73)

[12:51] Leia Dickerson: Also, I am a newbie to this field. Perhaps I could give a clarity check for definitions being created?     (2E74)

[12:51] John Sowa: And I'll be presenting the slides at a conference on knowledge graphs at Columbia in May.     (2E75)

[12:51] Ravi Sharma:     (2E76)

[12:51] John Sowa: That conference has now become a telecon.     (2E77)

[12:52] Mark Underwood @knowlengr: @John Good to know,,, would try to attend that     (2E78)

[12:52] Ravi Sharma:     (2E79)

[12:52] David Eddy: @Leia... yes... as a newbie, not pickled in this jargon, you would be a invaluable critic for clarity     (2E80)

[12:52] John Sowa: Another person who is speaking at the [tele]conference is Jans Aasman.     (2E81)

[12:52] janet singer: @David Actionable info and intelligent behavior are easily reconciled if they are framed in a general way that can span the cases of 1) machine 2) simple organism 3) human being and 4) combinations.     (2E82)

[12:53] Leia Dickerson: @David Eddy --thank you.     (2E83)

[12:53] BobbinTeegarden: @JohnS where can we find info about CLIP?     (2E84)

[12:54] John Sowa: Bobbin, I'll post a version of my slides and mention them on Ontolog Forum.     (2E85)

[12:55] Mark Underwood @knowlengr: @Bobbin - I will try to find a NEIM listserv, thx     (2E86)

[12:55] BobbinTeegarden: @JohnS thanks!     (2E87)

[12:56] John Sowa: But this is work in progress. Don't expect the progress for a while.     (2E88)

[12:56] BobbinTeegarden: @Mark if you go to OMG and look for FERM Meeting, the slides for the NEIM presentation will be posted (wait a few days)     (2E89)

[12:56] David Eddy: Since I'm a voice of one, I will be chief cook & bottle washer on "un-natural language"     (2E90)

[12:58] David Eddy: @Leia... one of my BIG windmills is there is NO professional cross fertilization between process of defining, creating, maintaining software & librarianship NONE.     (2E91)

[12:58] Ram D> Sriram: Ken: I think you will need to monitor the chat queue.     (2E92)

[12:59] janet singer: Leia mentioned     (2E93)

[12:59] Mark Underwood @knowlengr: @Leia - Can you post a link to that repo of KB repos?     (2E94)

[13:00] Leia Dickerson: @Mark-- Janet has the link above.     (2E95)

[13:00] David Eddy: gotta go...     (2E96)

[13:00] John Sowa: Mark, yes. Fortunately almost nobody who talks about knowledge graphs knows CLIP     (2E97)

[13:01] ToddSchneider: Some semi-random thoughts on what a knowledge graph is or maybe: 0) It's a model of a portion of some domain, 1) Is allows more fidelity of the representation of a domain via relations (the relations are the key), 2) The additional fidelity of representation (via relations) allows more more useful and actionable distinctions.     (2E98)

[13:02] BobbinTeegarden: @JohnS is CLIP a continuation of the knowledge tool (thought was CLIPS) from 80s?     (2E99)

[13:02] Gary Berg-Cross: Many different worlds. even larger than the many different words.     (2E100)

[13:02] Mark Underwood @knowlengr: @Leia Bunmmed... nothing in Bartoc for cybersecurity     (2E101)

[13:02] David Eddy: [return.. wrong top of the hour]     (2E102)

[13:02] David Eddy: Safe to say I'm only voice in favor of looking at "un-natural language?"     (2E103)

[13:03] Gary Berg-Cross: @Ram proposes some outline for the next week.     (2E104)

[13:03] John Sowa: Mark, I hit enter before I typed the S at the end of CLIPS. It's unlikely to cause confusion.     (2E105)

[13:04] Gary Berg-Cross: Ran volunteered to put thoughts together on standards for KG (DOL is the one).     (2E106)

[13:04] John Sowa: I have to go.     (2E107)

[13:05] David Eddy: required read: Craig Murphy & JoAnne Yates: Engineering Rules: Global Standard Setting since 1880"     (2E108)

[13:05] George Hurlburt: In the real world, most systems are dynamic. This means that data and metadata regularly come and go, morphing both systemic context and usage in the process. Even seemingly remote environmental variables serve to influence systemic context on the fly. Correspondingly, common definitions must morph equally fast. As can been seen in the daily news or among standardization efforts, getting consensus is the rub. Without such temporal consideration , however, static definitions only add limited value to understanding. COVID-19 is a good exemplar of such a morphing experience, although a highly exaggerated case of the real non-static world. To be useful, the KG/DB/Model must be situational in nature.     (2E109)

[13:05] janet singer: @John: Id like to work on the glossary. I do think we need some framing of the pragmatic cycle as the invariant context for all of these terms and concepts related to data, theories, models, inferences, evaluations > action     (2E110)

[13:05] Ram D> Sriram: @john: Are you referring to the production system language? I have implemented a version in C++ and Java in the early 1990s.     (2E111)

[13:05] janet singer: The pragmatic cycle is the only concrete standard we have     (2E112)

[13:06] Ravi Sharma: Paul - triples are foundation on which KGs are built, hope you agree but triples are not automatically converting them to a set of KGs what else except logic reasoning and other methods of linking triples and filtering them so their relevance to say a search topic is derived i.e. KGs?     (2E113)

[13:07] John Sowa: Ram, yes. But I'm using the word CLIP as a more readable notation for CLIF.     (2E114)

[13:08] John Sowa: Later today, I'll say more about models. See the note I wrote this morning.     (2E115)

[13:08] John Sowa: Have to go. Bye.     (2E116)

Resources     (2F)

Previous Meetings     (2G)

Next Meetings     (2H)