From OntologPSMW

Jump to: navigation, search
[ ]
Session Standards for Knowledge Graphs
Duration 1.5 hour
Date/Time 13 May 2020 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CEST
Convener Ram D. Sriram
Track What


Ontology Summit 2020 Standards for Knowledge Graphs     (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)

We have witnessed considerable activity in the development of knowledge graphs and networks. In order for these advancements to be widely used, we will need standards to ensure performance, conformity, and safety in applications. Further, standards play a very important role in innovation. Our panel will discuss various questions related to standards and knowledge graphs/networks – Why do we need standards? What is the state of the art? What are the future directions? What should various organizations be doing?     (2B1)

Conference Call Information     (2C)

Attendees     (2D)

Discussion     (2E)

Standards     (2E1)

[12:34] Ravi Sharma: Elisa - who are current users of SBVR which we worked on, as well as Date Time?     (2E1A)

[12:42] Gary Berg-Cross: Also from Ravi Ravi - with respect to SBVR, there is a community of consultants that have been using it for years, as well as a few companies - mainly for business modeling and streamlining business process rather than for, say, software development or other IT purposes; DTV is used with SBVR, but we have a subset of the ontology that Mark Linehan and I developed in FIBO     (2E1B)

[12:19] Ravi Sharma: John So DOL and CLIP are the KG Standards or are there optimized for use in KG?     (2E1C)

[12:21] Ravi Sharma: John- What is the role of Metadata in KG standards, standards need them or are they to aid in retrieval of search results, say from a larger collection or set?     (2E1D)

[12:36] John Sowa: Some of those issues are covered in more detail in the full set of slide:     (2E1E)

[12:27] Ravi Sharma: Elisa - Is UML2 or BPMN 2 now including graphics and does it aim at working with DOL?     (2E1F)

[12:40] Gary Berg-Cross: Ravi - UML2 and BPMN both include diagram notation, whereas DOL does not - it leverages the underlying model. Because UML and BPMN are much older and more widely used, DOL has really done the mapping rather than the other way around.     (2E1G)

[12:39] Igor Toujilov: Question to All: What is the place of the OWL 2 standard now?     (2E1H)

[12:43] John Sowa: Re OWL: All versions of OWL map to Common Logic -- in any dialect.     (2E1J)

[12:37] Mark Underwood @knowlengr: Elisa: What's up at ISO in this space?     (2E1K)

[12:43] Gary Berg-Cross: From @Mark - I haven't been involved in ISO work directly for a number of years, although I monitor some areas, including work coming out of SC32 for metadata, the vocabulary work, and a few others where we are planning to submit OMG standards via ISO PAS     (2E1L)

[12:45] Gary Berg-Cross: From @Mark - some work is being done to publish certain ISO Standards, such as the GLEIF LEI work, as OWL ontologies, with the resulting data sets published on     (2E1M)

[12:47] Mark Underwood @knowlengr: OK, @Elisa, got your pseudonym .. Is that via ISO WG42 ("AI") or SC32 data management?     (2E1N)

[12:51] Gary Berg-Cross: From @Mark - most of my work has been with JTC 1 SC32 data management     (2E1O)

[12:54] Mark Underwood @knowlengr: @Elisa - Thanks. Wish I could get Synchrony to join OMG     (2E1P)

[12:55] Mark Underwood @knowlengr: (I represent myself / Krypton Brothers in this venue)     (2E1Q)

[12:57] Elisa Kendall: @Mark - there will be opportunities later this year for folks to join managed communities that OMG enables with a much lower fee structure, so watch for that     (2E1R)

[12:58] Mark Underwood @knowlengr: @Elisa Good to know, will advise     (2E1S)

[12:59] Ravi Sharma: Elisa- that is good news, I had to leave OMG as I retired and could not pay high fees!     (2E1T)

[12:59] Mark Underwood @knowlengr: @Elisa Hope to nudge the Universal Cybersecurity Ontology (MITRE et al) toward standardization     (2E1U)

[12:59] Ravi Sharma: Elisa - ODM, SVBR Date and Time, Later UML profiles etc.     (2E1V)

Financial Ontologies     (2E2)

[12:48] Ravi Sharma: Mike- When will FIBO based applications and institutions will exchange contracts for CDOs and other securities?     (2E2A)

[12:51] Ravi Sharma: Mike- where the KG connection is strong in FIBO?     (2E2B)

[12:52] Mike Bennett: @Ravi there is extensive draft material on CDOs and CDO Square, along with CMO and so on, in the draft material in FIBO but not yet in the published OWL standard - it is part of that very conceptual stuff I mentioned and would need extraction on a use case basis (specific competency questions) for a given OWL based application.     (2E2C)

[12:57] Mike Bennett: @Elisa (probably) says: "@Ravi and @Mike - we paused on integrating the asset-backed security content in the FIBO working groups I co-chair until at least some part of the loan ontologies are more stable and published in released form, but they are on our roadmap in Securities once some of the loans work has been integrated?     (2E2D)

[13:02] Pete Rivett: GLEIF ontology for Legal Entities is outlined here . We'll shortly be publishing live datasets in conformant RDF on as previously mentioned. LMK if you have any questions or want to be notified when it goes live     (2E2E)

Knowledge Graph Foundation     (2E3)

[12:58] Mike Bennett: @Andreas does this maturity model compare with the one being developed in the Knowledge Graph Foundation?     (2E3A)

[13:04] Pete Rivett: FYI all, last week at the KG Conference we launched Enterprise Knowledge Graph Foundation     (2E3B)

[13:06] Andreas Blumauer: Mike, I don't know the maturity model being developed by the KG foundation     (2E3C)

[13:06] Andreas Blumauer: where can I find it?     (2E3D)

Ontologies and Taxonomies     (2E4)

[13:02] Janet Singer: Andreas said don't mix up ontologies and taxonomies too much ?     (2E4A)

[13:05] Andreas Blumauer: Janet, we constantly see ontologies and taxonomies strangely intertwined, they should be governed by different people/stakeholder     (2E4B)

[13:07] Mike Bennett: @Andreas what do you man by Taxonomy? Many people use that term to refer to a concept model (which need not be purely taxonomic) and if that's what you mean I strongly agree that governance is by a very different community from that for e.g. OWL ontologies for applications or KGs.     (2E4C)

[13:07] Janet Singer: @Andreas, How do you define the two that you have a sharp distinction between the two?     (2E4D)

[13:07] Andreas Blumauer: Taxonomies are typically based on SKOS(-XL) and represent the linguistic model of a domain     (2E4E)

[13:08] David Eddy: @Andreas... how many taxonomic domains in an enterprise?     (2E4F)

[13:09] Andreas Blumauer: David, depends on the enterprise     (2E4G)

[13:09] David Eddy: @Andreas... small, Fortune 350, 10000 people     (2E4H)

[13:10] Janet Singer: @Andreas, Are you using taxonomy for glossary then?     (2E4I)

[13:10] Andreas Blumauer: David, could you ask your question once again? I don't understand     (2E4J)

[13:10] Andreas Blumauer: Janet, no a glossary is less complex than a taxonomy     (2E4K)

[13:11] Andreas Blumauer: and a SKOS-based taxonomy can nicely play together with OWL-2 ontologies     (2E4L)

[13:11] David Eddy: @Andreas... I'm assuming there are multiple taxonomies across the various functions in an organization     (2E4M)

[13:11] Andreas Blumauer: @David, of course, an enterprise typically maintains dozens of taxonomies being linked and mapped to each other, based on several ontologies like FIBO     (2E4N)

[13:12] Andreas Blumauer: It's a matter of fact, that enterprises want to use structured and unstructured data all together.     (2E4O)

[13:12] David Eddy: @Andreas... since organizations tend to not know how many "applications" they have... I would assume taxonomies are equally unknown.     (2E4P)

[13:12] Andreas Blumauer: Graph-based NLP and text mining works best on taxonomies     (2E4Q)

[13:12] Mike Bennett: @Andreas et al - I presume in the use of SKOS for taxonomies, you are covering about 6 kinds of 'broader' v 'narrower' e.g. isA, parthood, containment and so on. In the formal ontology we limit this to the generalization (isA) relation and usually call that a taxonomy (hence the confusion on terminology).     (2E4R)

[13:17] Bobbin Teegarden: Isn't the main difference between a taxonomy and an ontology that a taxonomy is tree shaped, and an ontology is a (multidimensional) graph?     (2E4S)

[13:18] Andreas Blumauer: To be clear: it's not an EITHER/OR situation here: you have to use both     (2E4T)

[13:18] Andreas Blumauer: it's only because of costs being involved that you make a distinction and decompose the work to introduce a governance model which works for all stakeholders involved     (2E4U)

[13:19] Mike Bennett: @Bobbin a taxonomy need not be tree-shaped. Even for just generalization relations, you need to support multiple inheritance. The broader (SKOS) kind of taxonomy has even more kinds of relation so it won't be a tree anyway.     (2E4V)

[13:19] Andreas Blumauer: Trees, for example, are not poly-hierarchical, whereas taxonomies are.     (2E4W)

[13:20] Andreas Blumauer: a SKOS:Concept can be at the same time a FIBO:Contract etc.     (2E4X)

[13:20] John Sowa: Andreas, in any business, NL documents are the lingua franca of every department and division.     (2E4Y)

[13:22] Andreas Blumauer: John, our clients typically want to link their NL documents to structured data.     (2E4Z)

[13:23] John Sowa: Andreas, yes. See     (2E4AA)

[13:27] John Sowa: In the cogmem.pdf slides, please look at the example of legacy re-engineering.     (2E4AB)

[13:27] John Sowa: That application was a spectacular success story.     (2E4AC)

[13:28] David Eddy: @JFS... not that I knew it at the time... but in 1980 I worked in an insurance company with a well populated data dictionary. They had found 70 different terms for the core business concept "policy number."     (2E4AD)

[13:29] John Sowa: We tried to convince a major consulting firm of the advantage of that technology.     (2E4AE)

[13:29] John Sowa: Their answer: You converted a project that would require 40 person-years to one that would require 15 person weeks.     (2E4AF)

[13:30] John Sowa: They told us No!!!!!!!!!!!!!!!!!     (2E4AG)

Vagueness     (2E5)

[13:01] John Sowa: Fundamental principle: All models are wrong, but some are useful.     (2E5A)

[13:02] John Sowa: C. S. Peirce: It's easy to be certain. One only has to be *sufficiently vague*.     (2E5B)

[13:03] John Sowa: The greatest weakness of ontologies: they aren't sufficiently vague.     (2E5C)

[13:06] Bobbin Teegarden: @JohnS is 'vague' just a higher level of abstraction?     (2E5D)

[13:20] Janet Singer: @John, I think your single gem-like node for Ontology in your hexagon is misleading but I'm not sure what would be better ...     (2E5E)

[13:29] Bobbin Teegarden: @JohnS is 'vague' just a higher level of abstraction? -- Yes, that was my question also     (2E5F)

[13:30] Janet Singer: Vagueness in natural language involves a space of implicit and tacit connections. Higher level of abstraction in a machine representation is still precise     (2E5G)

[13:31] John Sowa: Bobbin, for 'vague' consider any definition in an ordinary dictionary.     (2E5H)

[13:31] John Sowa: Another word: underspecified.     (2E5I)

[13:33] Bobbin Teegarden: @JohnS yes, underspecified ... but necessarily more abstract?     (2E5J)

Abduction     (2E6)

[12:23] Ravi Sharma: John role of abduction?     (2E6A)

[12:37] John Sowa: Ravi, abduction is the only way to generate new ideas or theories.     (2E6B)

[12:38] John Sowa: Induction generalizes from examples. By generalizing, it loses information about individuals     (2E6C)

[12:39] John Sowa: Deduction cannot generate anything new. It just derives implications from the old info.     (2E6D)

[12:41] John Sowa: Abduction introduces something totally new -- hypothesis (AKA guess). But abduction must be tested by induction and deduction.     (2E6E)

Kübler-Ross     (2E7)

[12:58] Ravi Sharma: Andreas- the Kübler-Ross curves look different than Gartner Hype cycle?     (2E7A)

[13:04] Andreas Blumauer: Kübler-Ross curve is the foundation for Gartner Hype Cycle     (2E7B)

Miscellaneous     (2E8)

[12:02] Ravi Sharma: Thanks it looks a great panel.     (2E8A)

[12:05] Ravi Sharma: Elisa we used to be on OMG team together. welcome     (2E8B)

[12:05] Ravi Sharma: Steve - nice to see you     (2E8C)

[12:11] Marcia Zeng: John Sowa's slides     (2E8D)

These slides are a subset (with some modifications) of the slides in     (2E8E)

[12:13] Andreas Blumauer: In 1982, Knowledge Graphs were invented in the Netherlands. The theory of Knowledge Graphs was initiated by C. Hoede, a mathematician at the University of Twente, and F.N. Stokman, a mathematical sociologist at the University of Groningen. See:     (2E8F)

[13:31] Ravi Sharma: Ram - thanks for putting together a great panel     (2E8G)

[13:31] Ravi Sharma: Ken thanks     (2E8H)

[13:31] Andreas Blumauer: thanks everyone     (2E8I)

[13:32] Leia Dickerson: Thank you everyone. Great presentations!     (2E8J)

Resources     (2F)

Previous Meetings     (2G)

Next Meetings     (2H)