From OntologPSMW

Jump to: navigation, search
[ ]
Session Michael Uschold
Duration 1 hour
Date/Time 15 April 2020 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CEST
Convener KenBaclawski
Track Use Cases


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)

  • Michael Uschold Knowledge Graphs in Industry: Examples and Lessons Learned Slides Video Recording     (2B1)
    • ABSTRACT: We summarize a variety of projects in different industries where we have built ontology-backed knowledge graphs. They are driving an increasing number of production applications. Several examples are in the finance industry: resolution planning, records management, operational risk and commodities market pricing and analytics. Other examples are electrical devices and industry analysis. A major lesson learned is to coordinate various ontology efforts across a large enterprise to avoid recreating the very silos that semantic technology is designed to get rid of.     (2B1A)
    • BIO: Michael Uschold has thirty years’ experience in developing and transitioning semantic technology from academia to industry. He pioneered the field of ontology engineering, co-authoring the first paper and giving the first tutorial on the topic in 1995 in the UK.

      As a senior ontology consultant at Semantic Arts since October 2010, Michael trains and guides clients to better understand and leverage semantic technology using knowledge graphs. He has built commercial enterprise ontologies in finance, healthcare, legal research, commodities market, consumer products, electrical devices, manufacturing, and corporation registration. The ontologies are used to create knowledge graphs that drive production applications. His experience has been distilled and communicated in the book: “Demystifying OWL for the Enterprise”, published in 2018.

      During 2008-2009, Uschold worked at Reinvent on a team that developed a semantic advertising platform that substantially increased revenue. As a research scientist at Boeing from 1997-2008 he defined, led and participated in numerous projects applying semantic technology to enterprise challenges. He is a frequent invited speaker and panelist at national and international events, and serves on the editorial board of the Applied Ontology Journal. He received his Ph.D. in AI from Edinburgh University in 1991 and an MSc. from Rutgers University in Computer Science in 1982.     (2B1B)
  • Discussion     (2B2)

Conference Call Information     (2C)

Attendees     (2D)

Discussion     (2E)

Ravi Sharma: Michael - please describe database and ontology together?     (2E1)

Ravi Sharma: Michael - you have to identify that company A entity and B entities are having common vocabularies     (2E2)

Ravi Sharma: Is this done by process modeling or by reverse engineering of each company functions?     (2E3)

Mike Bennett: Michael what are your thoughts on Enterprise Ontology versus application use cases (competency questions) i.e. how do you make sure the CQs are at the level of an overall enterprise ontology not a specific solution use case?     (2E4)

Ravi Sharma: What you meant by requirements of KG?     (2E5)

Ravi Sharma: Todd- I meant for you the above.     (2E6)

Mike Bennett: Why are you treating Jurisdiction and Country as the same concept?     (2E7)

Mike Bennett: For example, since you are building iteratively, there may come a tie when the legal concept of a jurisdiction and the locative concept of a country turn out to need to be distinct after all. How do you manage 'breaking' changes like this?     (2E8)

Ravi Sharma: Michael - Would you say that all knowledge graphs need to have triples and ontologies?     (2E9)

Ravi Sharma: boldly speaking what else is there in ontologies beyond triples, logic or relating triples, occurrence of triples(frequency, type of relations, collection patterns, etc?)?     (2E10)

Mike Bennett: Ira Baxter needs to be told about this chat facility     (2E11)

Janet Singer: @Ravi: I think Michael did just say a KG is always backed by an ontology     (2E12)

Leia Dickerson: @Mike Bennett Agree with your comment about Jurisdiction and Country     (2E13)

Mike Bennett: Thanks Leia. The meta-level question is more interesting to me - how do they manage the potential disruption of such anti-commensense decisions (I assume it was a decision and not just the modeler being pushed around by the words that customers use - I've seen that too).     (2E14)

Bobbin Teegarden: MikeU: Is there a way to create an ontology of your sparql queries so you could determine dependencies, effects of changes more easily?     (2E15)

Ravi Sharma: Janet - yes but how implicit or details of how to include?     (2E16)

Ira Baxter: Promise to be good from now on :-}     (2E17)

Mike Bennett: Welcome @Ira!     (2E18)

Ravi Sharma: Glad Todd asked that Q     (2E19)

David Eddy: @Ira... you can also go to full name in Zoom chat... helps to be known     (2E20)

John Sowa: deja vu all over again. These issues were debated in detail in the DB world of the 1960s and 70s. The conceptual schema work of 1978 - 198 established solid strategies for handling them.     (2E21)

Mike Bennett: If you get the Top Level Ontology partitions right then subsequent changes (in OWL or DL) should all be additive. All this breaking is only needed if they have let the client push them around into conflating things that ought not to have been conflated.     (2E22)

John Sowa: Cyc relearned the wheel in the 1980s, and they had some very good answers in the 1990s.     (2E23)

Evan Wallace: Is the big concern for evolution, non-monotonic changes?     (2E24)

John Sowa: And this debate sounds like a throwback to the 1970s.     (2E25)

Ravi Sharma: ken - i would like to ask last but first provide Evan Wallace a chance?     (2E26)

David Eddy: @JFS... remember it took some 200 years to evolve from Newton's alchemy (in the simple 3D world) to acceptance of Mendeleev's periodic table.     (2E27)

Ira Baxter: Right. But unless somebody realizes tools that change the application code, you find that the effort to make changes by hand tends to stop anything but adding single columns.     (2E28)

David Eddy: @JFS... data & systems essentially exist in an N-dimenional world     (2E29)

John Sowa: Evan, the big concern is that nobody learns from the lessons of other people who faced the same issues and solved them.     (2E30)

Janet Singer: @Ravi: Right. Simplifying assumptions 1) A KG corresponds to a triple store database 2) An ontology is a (generic, conceptual) schema for the KG, which can be seen as implicit if not explicit     (2E31)

Ira Baxter: .. and Michael told about his manual evolution process. Seems like the perfect analog.     (2E32)

Mark Underwood @knowlengr: Did Michael U mention which graph database they're favoring?     (2E33)

Mark Underwood @knowlengr: @MikeB "conflaught" => distraught     (2E34)

Mike Bennett: @Mark indeed. I made the word up on the spot but I might keep it.     (2E35)

Ravi Sharma: Janet - thanks we are coming closer to express KGs bit by bit, thx     (2E36)

Mike Bennett: It does concern me that when you start with use cases it's easy to let the client push you around with their (necessarily) naive understanding of what you can or can't do with words. Then you see word-driven 'ontologies' which are perfectly right for stand-alone applications but not scalable.     (2E37)

David Eddy: Great rant John     (2E38)

David Eddy: spot on     (2E39)

Mark Underwood @knowlengr: @MikeB unless they start with FIBO or some other standard (e.g., building code IEEE std), even the seeding of client "Questions" a la Michael Gruninger doesn't circumvent that myopia     (2E40)

Evan Wallace: @janet Is this group narrowing on a def for KGs that includes only instance data and not a model to support that data (e.g. Ontology)?     (2E41)

Bobbin Teegarden: @JohnS Could you write a paper/presentation that summarizes the historical wisdom, relating to our current thinking? And is there a URL to the Google work on this? And thank you for pointing at DOL!     (2E42)

Janet Singer: @Mike: Maybe if ontology is demystified as a generic conceptual schema more attention can be focused on the question of anticipating what conceptual model will maintain appropriate genericity in future     (2E43)

Mike Bennett: @Mark FIBO itself has fallen subject to that same myopia, unfortunately. A by product of approaching ontology as a technology effort, I'm afraid.     (2E44)

Mike Bennett: @Janet absolutely. It's a separate science from the 'plumbing' of OWL etc. but the industry has yet to understand that you don't just hire the smartest technical plumber for the job.     (2E45)

Ravi Sharma: Janet and Mike - CDM is driven by BPM or functions performed by an enterprise     (2E46)

Ira Baxter: Well, have an underlying shared logic is great. Somebody has to bell the cat and build all these tools, or all we get is talk.     (2E47)

Ravi Sharma: I agree that we need to find what else beside CDM does ontology contain?     (2E48)

Mike Bennett: @Mark back to your point about the questions - absolutely you need a good framework within which to ask those, since subject matter experts aren't going to e.g. volunteer when something is a Relative / contextual thing, or address temporal contexts, or digital twins versus actual things.     (2E49)

Janet Singer: @Evan: No, just exploring questions of relating KG, ontology model, etc. Hopefully converging on simple, robust answers     (2E50)

Ravi Sharma: Evan I knew this? please include me in your Qs     (2E51)

John Sowa: Bobbin, I uploaded a draft of my slides to     (2E52)

John Sowa: Those are slides I'll present at the conference on knowledge graphs in May.     (2E53)

Bobbin Teegarden: Thank you, John!     (2E54)

Bobbin Teegarden: Which conference in May?     (2E55)

John Sowa: Evan, knowledge graphs are being used for both data and metadata     (2E56)

John Sowa: A schema is data about data.     (2E57)

John Sowa: An ontology is a systematic set of data about data.     (2E58)

Bobbin Teegarden: @JohnS which conference on knowledge graphs in May? URL?     (2E59)

John Sowa: Janet, this elephant has more than 7 sides.     (2E60)

John Sowa: Even 7 blind men can't discover all the sides.     (2E61)

John Sowa: But there is a lot that has been and can be done.     (2E62)

John Sowa: See the KG2LL.pdf slides     (2E63)

John Sowa: Bobbin, I'll send a note to O forum this afternoon.     (2E64)

Ravi Sharma: Janet- in answer to Evan's Q we are wanting to do more than tie to a specific instance of KGs?     (2E65)

Resources     (2F)

Previous Meetings     (2G)

Next Meetings     (2H)