Actions

Ontolog Forum

Session Introductory
Duration 1 hour
Date/Time October 16 2019 16:00 GMT
9:00am PDT/12:00pm EDT
5:00pm BST/6:00pm CEST
Convener Ken Baclawski

Ontology Summit 2020 Proposals for Overview of Summit

Agenda

Conference Call Information

  • Date: Wednesday, 16-October-2019
  • Start Time: 9:00am PDT / 12:00pm EDT / 6:00pm CEST / 5:00pm BST / 1600 UTC
  • Expected Call Duration: 1 hour
  • The Video Conference URL is https://zoom.us/j/689971575
    • iPhone one-tap :
      • US: +16699006833,,689971575# or +16465588665,,689971575#
    • Telephone:
      • Dial(for higher quality, dial a number based on your current location): US: +1 669 900 6833 or +1 646 558 8665
      • Meeting ID: 689 971 575
      • International numbers available: https://zoom.us/u/Iuuiouo
  • Chat Room

Participants

Proceedings

[12:17] TerryLongstreth: John says (I paraphrase, I hope): "Knowledge graphs are at the level of ordinary language. " I assume the 'graph' part reflects the way that the linguistic tokens are tied together to capture detail beyond the language.

[12:20] RaviSharma: Terry are you talking of executable English

[12:21] Douglas R. Miles: I think that is giving too much ownership of "Knowledge Graph"

[12:22] Douglas R. Miles: I think John is giving too much ownership of "Knowledge Graph" to actors that are not interested in Logic

[12:22] Douglas R. Miles: "Bad Actors"

[12:22] Gary: We might think of Jano & Pascal's "purpose driven ontologies" as a version of micro-theories (MTs). They certainly are often "modular" such as [1]

[12:23] Douglas R. Miles: John's Knowledge Graphs (CGIF) are not Linguistically Derived

[12:25] TerryLongstreth: Ravi, I was trying to tie John's statement to our theme. I don't think executable English is a particularly typical example of system for producing knowledge graphs. I'm really trying to understand if TBX (Term base exchange) from ISO TC37 is another candidate example.

[12:41] Gary: You can see an OWL ontology used as as part of a personalize health KG(paper at [2]) at [3]

[12:43] TerryLongstreth: Fuzzy, vague, ambiguous, abstract - all different words with different shades of meaning.

[12:44] TerryLongstreth: -in contemporary technical English

[12:45] ToddSchneider: We should not confuse the notion of 'precision', as used in querying of information systems, with 'interpretation' (i.e., how data may be introduced into an information system or a knowledge graph).

[12:46] Gary: Just because we have a technique for fuzzy (fuzzy logic) does not mean it will address all the issues around ambiguity....as John has written on at length. And our summit on context exposed some thinking on this issue.

[12:46] RaviSharma: Kathryn Blackmond Laskey. search uncertainty ontologies please.

[12:46] anneogborn: toughness is the word he's looking for

[12:49] janet singer: Value of vague/robust vs precise/fragile value is relative to stability of context - gets at dynamics issue

[12:49] Gary: Kathryn Blackmond Laskey is at GMU

[12:51] janet singer: There is also precision in link statistics available in KGs, just not conceptually generalized precision as in a formal higher ontology

[12:53] Douglas R. Miles: take these to assertions.... Joe->loves->Kathy Kathy->loves->Joe these versions of Loves are not logically even close to the same things.

[12:54] Douglas R. Miles: loves(joe,kathy). loves(kathy,joe). .. none of these 6 atoms here are the same thing

[12:54] Douglas R. Miles: (Prolog-ish atoms)

[12:55] ToddSchneider: Kathy Laskey is on the faculty of George Mason University (Fairfax, VA)

[12:56] Douglas R. Miles: But, we can say (thereExists ?L1, ?L2 (and (subProperty ?L1 Love)(subProperty ?L2 Love) (?L1 Joe Kathy) (?L2 Kathy Joe )))

[12:57] janet singer: The issue is constraints on generalization

[12:57] Douglas R. Miles: Merely attempting to throw a bone (towards John) secretly our formal logics are "Signs" and "Mentions"

[12:58] Douglas R. Miles: Relations always are a subRelation of their name

[12:59] Douglas R. Miles: Instances always coerce themselves to being what aspects lets them fit into the Arg places

[12:59] Douglas R. Miles: (knows Joe Sally)  ;; these two Instances are not the same "type"

[13:00] Douglas R. Miles: Joe is a Knower type... Sally is a Knowable type

[13:01] Douglas R. Miles: (knows Sally Joe) (knows Joe Sally) .. Very likely the two Joe mentions are not at all about the same aspect of Joe

[13:02] Paul Brown: Also sounds like Piaget's ideas of schemas and moving towards equilibrium from constructivist pedagogy

[13:03] TerryLongstreth: I have to leave, but this has been one of the most valuable sessions for me personally. My hot button is the idea that everything we know is subject to change, so we have to provide change-sensitivity in our systems. We don't know what we don't know, but perhaps we can create systems that can recognize ideas that have not been previously encountered.

[13:04] RaviSharma: Ravi's notes: John said that the value of results from KG processing is not in a single triple match but in the fact that patterns of several triples and chains are matched.

[13:07] Gary: Albus was chief of the Intelligent Systems Division at NIST...Ram is chief there now.

[13:07] Douglas R. Miles: Since logic is considered to always be well formed we can say that "generalizations" will always constrain to the sanest possible truth

[13:07] janet singer: Yes re Piaget

[13:07] Douglas R. Miles: (Saying Symbols are getting Fatter/Skinner every few seconds and with context)

[13:08] Douglas R. Miles: Skinnier.. Skinner doesn't belong in any AI related discussion

[13:09] Paul Brown: Right? These theories seem to be all building on cognitive psychology. Piaget's Assimilation, accommodation and equilibrium are the belief revision step

[13:09] Paul Brown: We'll leave Skinner to the ML folks...

[13:13] Paul Brown: But this diagram is really interesting, it suggests how the constructivist and behaviourist schools fit together.

[13:14] Douglas R. Miles: Yes.. these cognitive theories translate well to a *Aristotelian* view of logic

[13:14] Paul Brown: Which is handy for ontology...

[13:14] RaviSharma: John said the power of Vivomind was in its ability to match these patterns gave example of legacy systems analysis.

[13:16] Douglas R. Miles: "*Aristotelian* view of logic" i should say *Aristotelian* view of reasoning .. an argument with one self that results in motive/motions

[13:16] janet singer: Data is always precise - ambiguity is in interpretation (as Todd said above)

[13:16] JimDisbrow: If we cannot express semantically that continuity over time exists in our functional relationships, rather than categories of time (past, present, future, conditional subjunctive, etc.) - and if we cannot exclude "is" and "isa" which force categories over precise relationships, movement toward integration of ontologies will stall. For example: How might the concept that parents' love of their children began before birth, lasts a lifetime and could possibly last long after the child's early death - how might this kind of verb action-relationship be installed in any English-based semantic ontology?

[13:17] RaviSharma: John says vagueness of KGs is useful for expected results

[13:18] BobbinTeegarden: Where can we find John's diagrams?

[13:20] janet singer: On the meeting page

[13:20] ToddSchneider: Bobbin, John's slides are available from today's meeting page.

[13:21] Douglas R. Miles: (isa Snoopy-555 Dog) (isa Buffy-666 Dog) .. They are indeed instances of Dogs.. But in a Query of (isa Buffy-666 ?WHAT1) and (isa Snoopy-555 ?WHAT2).. the WHAT1 and WHAT2 cannot be the same.. the have to have a "Circumscriptive Sameness"

[13:21] ToddSchneider: Today's meeting page URL: [4]

[13:21] Douglas R. Miles: They can circumscribe into a Dog

[13:22] Douglas R. Miles: Thank you Ken!

[13:23] Douglas R. Miles: This is hard for us all to accept because so much works when we use oversimplified methods of unification

[13:28] ToddSchneider: Meeting ends 13:25 EDT

Resources

Previous Meetings

... further results

Next Meetings

... further results