From OntologPSMW

Jump to: navigation, search
[ ]
Session Communiqué
Duration 1 hour
Date/Time 17 June 2020 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CEST
Convener KenBaclawski


Ontology Summit 2020 Communiqué Development     (2)

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)

Proceedings     (2E)

[12:16] Mike Bennett: Add data integration to the list of reasons to use KGs?     (2E1)

[12:16] David Eddy: re: NLP support... I would, of course, argue for recognition of the messy language inside software applications     (2E2)

[12:21] David Eddy: "cobbled together over decades" is more appropriate     (2E3)

[12:22] John Sowa: Knowledge graphs are limited in expressive power to RDF or RDFS     (2E4)

[12:25] Todd Schneider: Is the constraint on expressivity of Knowledge Graphs due to the restriction to only unary and binary relations?     (2E5)

[12:28] Ravi Sharma: John Sowa said that KGs use at most the power equivalent to RDF (RDFS) and Ontologies are providing the Terms in Knowledge Graph.     (2E6)

[12:29] Ravi Sharma: DOL is foundation for KG (using UML) first standard for KGs.     (2E7)

[12:30] Ravi Sharma: Gary said use of OWL     (2E8)

[12:33] Gary Berg-Cross: I showed a slide from Cogan's presentation on Comprehensive Modular see Ontology Engineering     (2E10)

[12:34] John Sowa: Please look at     (2E11)

[12:39] Bobbin Teegarden: Can you create a knowledge graph in UML?     (2E12)

[12:40] Ravi Sharma: John referenced DBPedia and KB Pedia which relate to KG implementation, originally Google coined word KG for semantic network     (2E13)

[12:41] Ravi Sharma: Invite Kingsley     (2E14)

[12:43] John Sowa: To show how all the logics are interrelated, I recommend that diagram from the DOL standard.     (2E15)

[12:43] Ram D. Sriram: @john: I presume someone has Kingsley's contact information     (2E16)

[12:44] Ken Baclawski: Proposed outline (by Todd Schneider):     (2E17)

  • Introduction     (2E18)
  • Whence - what brought the use of graphs in as persistence mechanisms (will need to address other non-relational persistence mechanism); historical background; this can include parts of 'why'     (2E19)
  • Current State - How are graph persistence mechanisms being used     (2E20)
  • Problems - what are the differences in the current uses and what problems they may cause going forward     (2E21)
  • Whither - Our recommendation of how a the notion 'knowledge graph' should be defined; How standards can help (and maybe a lead in to the next year's summit topic).     (2E22)

[12:44] John Sowa: It shows how every one of those SW and fUML notations are related to Common Logic.     (2E23)

[12:45] Bobbin Teegarden: Are various UML graphs KGs? Class diagram (graph), activity diagram (graph), even sequence diagram (graph)...?     (2E24)

[12:47] David Eddy: what is difference between "repository" & database?     (2E25)

[12:48] David Eddy: I want to collect it because I just bought a 4TB disk drive for $100. I've got to fill it with something... /sarc     (2E26)

[12:49] Gary Berg-Cross: Comment on Ravi's slide - we should separate "what is knowledge?" and "how is it represented" (per Mark Musen's talk).     (2E27)

[12:50] John Sowa: By the way, my talk at the KG conference in May was awarded the best presentation award.     (2E28)

[12:51] John Sowa: The ESCW slides were presented as a one-hour keynote talk at the European Semantic Web Conference.     (2E29)

[12:52] Alex Shkotin: @John, congrats!     (2E30)

[12:52] Janet Singer: @Gary: and Mark Musen got into the Why question raised by Todd     (2E31)

[12:52] Bobbin Teegarden: URL to Mark Musen's talk?     (2E32)

[12:54] Janet Singer: Musen is the second speaker here     (2E33)

[12:54] John Sowa: Alex, thanks for the congrats. But I emphasize that those slides are *not* a summary of my own work.     (2E34)

[12:54] Bobbin Teegarden: I think a graph is a collection of RELATED triples, no?     (2E35)

[12:54] John Sowa: They are a summary of what people have been doing with KGs over since Google introduced the term in 2012.     (2E36)

[12:55] Bobbin Teegarden: Thanks, Janet!     (2E38)

[12:56] John Sowa: Bobbin, it's possible to map any graph to triples. But that does not mean that the triples are the best way to store, process, or think about the data.     (2E39)

[12:56] Gary Berg-Cross: Note, we can cite more recent KG work than Google graphs for search. There are financial, healthcare and product KBs. Brad Bebee talked about Putting data into context with Amazon Neptune. New KGs on corona virus etc. as current work.     (2E40)

[12:57] John Sowa: Gary, yes. Much more work has been done. But I repeat, the eswc.pdf slides shows the *foundations* for all that work.     (2E41)

[12:58] Gary Berg-Cross: As JPMorgan Chase has a relationship with ~50% of U.S households and >80% of Fortune 500 companies, its global data footprint presents significant opportunities. Although Knowledge Graphs were traditionally limited to Fraud and Risk management, recent usage is growing across several areas including AI-driven customer care, generating alpha from alternative data, etc. The Company Knowledge Graph developed by combining internal client data with third party licensed data can be instrumental in answering a wide variety of questions ranging from identifying weak links in the supply chain to investments in startups. Commonly faced challenges with Knowledge Graphs arise from extracting insights from structured and unstructured data, dynamic nature of financial information and required access control mechanisms. The industry and academia see significant potential in future Knowledge Graph applications around Natural Language Understanding, Question & Answer systems, Recommender systems, and Reasoning.     (2E42)

[12:59] Gary Berg-Cross: Intuit, the leading financial software/service company behind TurboTax, Mint and Quickbooks, is embarking on a multi-year transformational journey into an AI-driven Expert Platform to help Small Business, Self-Employed and Consumers to prosper. The key pillars of the platform rely heavily on clean data and intelligent systems. In this talk I will share two specific knowledge graph use cases in data integration and tax logic programming that helped Intuit solve big business and customer problems at scale.     (2E43)

[12:59] John Sowa: Gary, those are applications. But the logic determines what can and *cannot* be done.     (2E44)

[13:01] David Eddy: I'll volunteer to deal with legacy systems... knowing full well, no one else is interested.     (2E45)

[13:01] Ram D. Sriram: I can do work on the last one -- probably standards     (2E46)

[13:01] Gary Berg-Cross: I will be involved in the Problems piece...but also can add to the current state on KG engineering practices.     (2E47)

[13:02] Todd Schneider: I can work on 'Whither'.     (2E48)

[13:02] Alex Shkotin: 'Whither' if I can.     (2E49)

[13:03] Janet Singer: John, I agree that we should use DOL as part of clarifying the KG and Ontology relationships. Are #3 and #4 the characterizations you give for those two in your slides? 3. Symbolic models consist of words related by words to other words. 4. Ontology is a catalog of words and the kinds of things they refer to     (2E50)

[13:03] David Eddy: Current State... does it support documenting operational, legacy systems     (2E51)

[13:06] David Eddy: @Ram... yes please... need help     (2E52)

[13:06] Gary Berg-Cross: One of the future directions topic that the Stanford course included was explanations - systems get to explains how it came up with a recommendation using a KG for reasoning.     (2E53)

[13:07] Gary Berg-Cross: Another direction is systems that engage like intelligent assistants ...and include commonsense.     (2E54)

[13:07] Doug Holmes: I'll join the Whence group     (2E55)

[13:08] Janet Singer: In whence we can cover 1) Representations, 2) Logics, 3) Knowledge     (2E56)

[13:08] Mike Bennett: I can't join any groups next week as we have the OMG meetings.     (2E57)

[13:09] Janet Singer: I volunteered myself and John for Whence as well     (2E58)

[13:22] Alex Shkotin: @Todd what about Whither group meeting and plans?     (2E59)

[13:28] Todd Schneider: Alex, depending on the outline we my need to wait to see what is written in the other sections.     (2E60)

[13:31] Alex Shkotin: Todd, we may continue this topic by our google group?     (2E61)

[13:37] Alex Shkotin: John, if symbolic model includes numbers it is not very symbolic.     (2E62)

[13:45] Janet Singer: J Sowa said KG starts from the data, and you can derive the theory; ontology is theory that can be applied to the data     (2E63)

[13:45] Janet Singer: KG or symbolic model is a descriptive model     (2E64)

[13:47] Janet Singer: KG talks about instances     (2E65)

[13:47] Alex Shkotin: @Janet, we should add that theory is formal and model is finite and is a model of this theory in Tarski style.     (2E66)

[13:48] Janet Singer: Ontology is the generic, describes the types     (2E67)

[13:49] Alex Shkotin: Sorry, need to go     (2E68)

[13:49] Janet Singer: Thx for the point     (2E69)

[13:51] Janet Singer: I think we need to clearly address the theory vs model and theory vs data relationships     (2E70)

[14:02] Janet Singer: Ontology is the theory - it can start from math, be bottom-up, etc. Ontology captures the definitions, think of it like a dictionary     (2E71)

[14:03] Janet Singer: But ontology refers to the deep structure of the distinctions in a domain, whether captured in a dictionary yet or not?     (2E72)

[14:07] Janet Singer: The world is dynamic, our knowledge is dynamic, our explicit representations of theories need to be dynamic     (2E73)

[14:12] Ravi Sharma: part of discussion with John, Janet, george - john says -mathematical, actual and laws of nature are part of top level ontology.ontology     (2E74)

[14:13] Janet Singer: Divide top level into: possibilities and necessary conclusions in mathematics (including everything imaginable); actuality; the laws governing (nature and science, which approximate laws of nature)     (2E75)

[14:14] Ravi Sharma: George said - perception vs actuality - approximations for a purpose (ravi)     (2E76)

[14:15] Ravi Sharma: John said pure math are indep of anything but analysis and reasoning relate to actuality.     (2E77)

[14:15] Ravi Sharma: Geometry etc are used probability to describe actuality.     (2E78)

[14:18] Janet Singer: Pure math unconstrained by actuality, applied math relates law/theory and actuality with pure math     (2E79)

[14:18] Ravi Sharma: math is required to be consistent - John     (2E80)

[14:20] Ravi Sharma: possibility actuality and necessity - janet and john     (2E81)

[14:21] Ravi Sharma: john - it is a piercian triad     (2E82)

[14:26] Ravi Sharma: cognition, perception, psychology to neuroscience, KNOWING     (2E83)

[14:27] Ravi Sharma: ontology distinctions and decision?     (2E84)

[14:28] Ravi Sharma: janet - informed decision     (2E85)

[14:28] Ravi Sharma: conceptual schema is better term than ontology - John     (2E86)

[14:30] Ravi Sharma: George - DB schemas need redesign?     (2E87)

[14:34] Janet Singer: Graph orientation allows conceptual, logical and physical models to be in greater alignment     (2E88)

Resources     (2F)

Previous Meetings     (2G)

Next Meetings     (2H)