Chat transcript from room: summit_20151210
2015-12-10 GMT-08:00
[09:25] MichaelGruninger: Agenda:
1. Discussion on scope of the theme -- Semantic Integration
2. Discussion on structure of the Summit
3. Identify Tracks
4. Nominate Chairs
5. Nominate Track Champions
6. Schedule Launch session (January 2016)
7. Schedule Tracks?
[09:30] anonymous morphed into ChristopherSpottiswoode
[09:32] anonymous morphed into Gary Berg-Cross
[09:32] anonymous1 morphed into AnatolyLevenchuk
[09:36] anonymous morphed into Judith Gelernter
[09:38] anonymous morphed into BethDiGiulian
[09:39] ToddSchneider: I'm having trouble
accessing Skype. Anyone else experiencing the same problem?
[09:40] Gary Berg-Cross: @ToddSchneider 12:39] I
had trouble before 12:30 and it took 4 tries.
[09:40] Uri Shani (IBM): same here
[09:41] ToddSchneider: I'm trying from a
different machine and OS.
[09:41] ChristopherSpottiswoode: Same here too.
Skype is shedding tears onscreen!
[09:41] Donna Fritzsche: Discuss overall
timing/schedule. Very intense marathon in winter-spring.
[09:41] Mark Underwood: RE Skype Took a couple
of tries, but I'm on it now
[09:42] MichaelGruninger: **** 1. Discussion on
scope of the theme -- Semantic Integration ***
[09:42] TerryLongstreth: We started out as 4 or
5 day FtoF conferences. AFAIK we went to the 3 months of weekly
sessions to try to cover the same ground.
[09:42] anonymous morphed into Veda Storey
[09:43] MichaelGruninger: Focal questions we
would expect to have addressed in each session:
What are the semantic integration challenges that are being
faced? How are ontologies expected to play a role? What new
techniques are needed? How are existing techniques working or
failing?
[09:45] MichaelGruninger: Potential Track topics
raised at the Brainstorming II session
[09:45] MichaelGruninger: Topic 1: Financial
Industry -- MikeBennett, John Yanosy
Topic 2: Earth Sciences Gary Berg-Cross
Topic 3: Biomedical Donna Fritzsche
Topic 4: Cloud Services Mark Underwood
Topic 5: Sensor Fusion Ken Baclawski
Topic 6: Manufacturing/Engineering
Topic 7: Cybersecurity
[09:45] MichaelGruninger: GaryBerg-Cross: Scope
shouold included semantic interoperability / data
interoperability / data sharing
[09:47] MichaelGruninger: What ideas are
included under "semantic integration"?
[09:48] ToddSchneider: Semantic Mediation
[09:48] ToddSchneider: Semantic Architecture
[09:48] MichaelGruninger: GaryBergCross: Focus
is on the interoperability / integration of systems rather than
restricting integration to ontologies per se
[09:48] ToddSchneider: Ontology Architecture
[09:49] Uri Shani (IBM): My interpretation:
Semantic mediation would be integrating models based on
ontologies, not integrating the ontologies themselves.
[09:49] ToddSchneider: Agree with Gary.
[09:50] Donna Fritzsche: Look at broad umberella
of ideas relating to sematic interoperability and architecture
[09:50] AnatolyLevenchuk: Usage of embeddings
and distributed semantic representation to map ontologies (how
deep learning can help ontologies and vice versa). Se
https://www.google.com/search?q=distributed+representations
[09:51] TerryLongstreth: Situations where
semantic integration needs arise: 1) legacy systems with a) no
ontology conceptions
b) incompatible NL bases (where semantics are aspects of human
language
2)systems where current Ontology tools are inadequate and
require/d procedural supplementation
[09:51] ToddSchneider: Uri, semantic integration
could involve both.
[09:52] Donna Fritzsche: ideas, solutions, best
practices, failings, shortcomings
[09:52] MichaelGruninger: Although ontology
integration is not the focus or motivation, it will definitely
be part of the solution in a scenario in which each software
application has its own ontology
[09:52] Gary Berg-Cross: I agree with Mike (and
Uschold's) point that withing some system attempts to integrate
a given system may have its own, local ontology and thus to
integrate more broadly with other systems having their own
ontologies, one must then also "integrate" (or bridge) these
ontologies.
[09:52] Mark Underwood: Agree that we would not
want to exclude the scenarios Terry mentions - Would be common
with cloud service interop undertaken by non-AI non-KBS folk
[09:54] Uri Shani (IBM): Gary, again in my
interpretation, semantic mediation would in fact "bridge"
ontologies.
[09:55] Donna Fritzsche: see what
solutions/architecture arise out of different domain areas (per
Mike)
[09:55] ToddSchneider: The notion of 'integrate'
suggests a static one time solution. Whereas 'mediation'
suggests (I think) something more flexible.
[09:56] Donna Fritzsche: True - to Todd - does
semantic architecture encompass both?
[09:56] Uri Shani (IBM): The common use case
would be: data (model) for certain ontology is mediated and can
be inerpreted according to another ontology.
[09:56] Donna Fritzsche: architectures that
integrate, mediate, interpret
[09:56] ToddSchneider: Donna, certainly, but it
depends on the problem to be solved (and the client's
willingness:)
[09:56] Donna Fritzsche: and faciliate
[09:56] Uri Shani (IBM): nice
[09:56] Mark Underwood: The synthesis advocate
can be the loose rein on topic drift
[09:57] Donna Fritzsche: to Todd - that is the
point of covering different domains - to tease out these
variations.
[09:58] Uri Shani (IBM): my examples of semantic
mediation are in the systems engineering. That would be tools
interoperability through that.
[09:58] TerryLongstreth: Donna, Todd, ---- and
find /exploit similarities/differences
[09:58] MichaelGruninger: We could also have a
series of sessions in which the Applied Ontology community
can then respond with potential solutions to the problems raised
by the Domain Tracks
The first synthesis session would summarize the challenges which
the Applied Ontology
community would address and the second synthesis session would
determine how
well the proposed solutions meet the challenges.
[09:59] AnatolyLevenchuk: There are different
means now to express ontology and semantics, not only logic.
See, e.g.
http://msrvideo.vo.msecnd.net/rmcvideos/258666/dl/258666.pdf --
then you can use map two ontologies/data models to embeddings
space that you can get from deep learning techniques. This is a
challenge, to use non-logical representations of ontologies!
[09:59] Judith Gelernter: Could representatives
from the domains volunteer data sets/databases that require
integrating? These might be used for case studies.
[10:00] Donna Fritzsche: at a meta level - we
can see differences and similarities more clearly as they relate
to use cases - time, granularity, accuracy, cost. We can start
to understand which models work in which use cases/domains
[10:00] Gary Berg-Cross: I agree that one or
more synthesis sessions is an important step leading to the
Symposium.
[10:02] Mark Underwood: Judith - Agree - the
track mgr should try to attract prospective datasets, though I
see from the Big Data CoI that one often gets vague description
vs schema dumps; i.e., compelling use cases, but with
generalized descriptions of data elements
[10:03] Gary Berg-Cross: It is my experience
that within the Earth Science and the Biomedical area ontologies
are increasingly being looked at as ways of going beyond domain
controlled vocabularies and related metadata effort. I would
expect to have issues around this discussed in one of more these
domain sessions.
[10:03] ToddSchneider: Most drivers of semantic
integration involve a need to cross a boundary (of some sort)
that hadn't been crossed or needed to be crossed before. So
perhaps the tracks could look for examples of these sorts.
[10:03] Uri Shani (IBM): 7 is good number... :)
[10:04] Donna Fritzsche: Todd - there are some
solutions that start to accomplish this
[10:04] Uri Shani (IBM): I can help with #6 - my
domain.
[10:05] ToddSchneider: Looking at data/data sets
in isolation won't help to understand nor convey the value of
semantic integration. The larger context of why the data needs
to be shared or accessed (across boundaries) provides the
motivation.
[10:06] TerryLongstreth: +1 Todd
[10:06] ToddSchneider: P.S. Skype is a total
failure for me today. So I'm only responding to what's in the
chat.
[10:07] Donna Fritzsche: bummer todd!
[10:09] Donna Fritzsche: good point Tood - on
data sets comment - use cases/context/requirements/need drive
the integration and architecture needs
[10:10] MichaelGruninger: Confirmed Tracks:
[10:10] MichaelGruninger: 1. Earth Sciences
(Gary Berg-Cross, KenBaclawski)
[10:10] MichaelGruninger: 2. Biomedical (Donna
Fritzsche)
[10:11] Donna Fritzsche: Music community is
doing alot.
[10:11] Donna Fritzsche: (re cloud)
[10:12] Donna Fritzsche: IBM Bluemix / Watson
comments
[10:12] MichaelGruninger: 3. Cloud Services
(Mark Underwood)
[10:15] MichaelGruninger: 4. Systems Engineering
(including manufacturing)
[10:18] AnatolyLevenchuk: ISO 15926 community
still working on their solutions in engineering. But there are
no huge success still with it. All CAD/PLM vendors declare
support of this ontology, but in reality there are no real
interoperability yet. Too much manual labor to map 'em all!!!
[10:18] Donna Fritzsche: ISA-88 and 95 in
manufacturing is a related set of efforts
[10:19] TerryLongstreth: @Donna - can you
elaborate on music community (machine composing, editing,
synthesis ....?)
[10:19] Mark Underwood: RE Cloud services: What
I was trying to articulate about API-driven design was better
explained here:
http://techcrunch.com/2015/09/27/the-future-of-coding-is-here-and-threatens-to-wipe-out-everything-in-its-path/?ncid=rss&cps=gravity_1730_-931760535667349867#.vrjnuun:y8he
[10:20] AnatolyLevenchuk: System engineering
fade in its ontology discussians. In INCOSE there are not rise
of ontology mentions.
[10:20] Donna Fritzsche: for music - curation -
just based on what I have seen in the news and job boards
[10:20] LeoObrst: See Ontology Summit 2012:
Ontology for Big Systems:
http://ontolog.cim3.net/OntologySummit/2012/. For systems
engineering, etc.
[10:21] Uri Shani (IBM): Anatoly - indeed.
[10:21] Uri Shani (IBM): I made the point of
tools interoperability for SE tools.
[10:21] ToddSchneider: Anatoly, INCOSE is not
known for being forward thinking.
[10:21] AnatolyLevenchuk: New ontologies in
clouds are services of image recognition (there should be some
kind of ontology to classify images and video). E.g.
http://clarifai.com/
[10:21] Uri Shani (IBM): :)
[10:21] MichaelGruninger: NIST has offered April
19/20 for the Symposium
[10:22] AnatolyLevenchuk: @Todd. Yes )))
[10:22] MichaelGruninger: One possible schedule:
[10:22] MichaelGruninger: January
14 Launch Session
21 Track X
28 Track Y
February
4 Track Z
11Track W
18 Track V
25 Synthesis I
March
3 Potential ontology-based solutions I
10 Potential ontology-based solutions II
17 Synthesis II
24 Communique I
31Communique II
April
7
14 19/20 Symposium
[10:22] Donna Fritzsche: cool - Anatoly
[10:22] Judith Gelernter: I suggest limiting
each session to 1 hour.
[10:22] Uri Shani (IBM): Although INCOSE honored
us with best paper a couple of years back on a paper promoting
semanatic mediation and ontologies for model-based SE tools.
[10:24] Mark Underwood: Anatoly - thx for
Clarifai link
[10:26] TerryLongstreth: Major difference in
recent Symposia is the requirement to have a completed
communique before the FtoF.
[10:28] AnatolyLevenchuk: Clarifai is only one
of many, but this is more about it than simply "cloud ontology".
Please, think about non-logical ontology representation (e.g. to
compare image with ontology type there should be common
representation for images and types!). This is representation
learning, deep learning most prominent. You should invite
somebody from that community. It is most disruption technology
from all up to now.
[10:29] AnatolyLevenchuk: My point is that
computational ontology not bounded in 2015 to logic-based
representations. It should be somehow reflected by agenda of a
Summit.
[10:30] Donna Fritzsche: +1 anatoly's point
[10:30] Judith Gelernter: Quality of output
would be higher if the sessions were more succinct.
[10:30] AnatolyLevenchuk: Ontology is about
"what is in the world", answers up to now was logic-like, but
today answers can be neural-like.
[10:32] Donna Fritzsche: Creating more succinct
sessions - requires more time input up front
[10:33] Judith Gelernter: Yes, and it would be
worthwhile
[10:34] Mark Underwood: For me, the level of
effort is the burden, not the elapsed time, but that may be a
subjective preference
[10:36] Judith Gelernter: To keep the sessions
to 1 hour, what would be required would be to limit the number
of slides per speaker (to say, 15).
[10:36] Donna Fritzsche: what do you mean mark?
can you talk
[10:36] Donna Fritzsche: and targeted questions
- @judith
[10:38] LeoObrst: Have to go, folks. Thanks.
[10:38] Mark Underwood: Ciao, Leo
[10:38] AnatolyLevenchuk: Ontology of style and
non-word-expressable objects. E.g. semantic morphying (query of
objects that you have no keywords) --
http://illustration2vec.net/ (this is about possibility mapping
of two ontologies in common semantic space that is
multidimentional meanings space, not ontology graph).
[10:38] Donna Fritzsche: agree Judith!
[10:39] Judith Gelernter: We will get better
slides -- and better presentations -- if we require speakers to
think about what they want to say, and limit their slides.
[10:39] Gary Berg-Cross: On this topic of
speaker time for the sessions. We have had 2 hr sessions with
the last half hour for disussion. This allowed 3-4 speakers per
session.
[10:41] Donna Fritzsche: What is the value of so
many tracks? maybe we need to think on that?
[10:41] Judith Gelernter: Sure. We can restrict
the number of tracks.
[10:41] Donna Fritzsche: What areas is doing the
best work or has the most interesting open questions?
[10:41] TerryLongstreth: One track could be
cross track lessons to be communicated
[10:42] TerryLongstreth: Commitment is a problem
[10:42] Judith Gelernter: Speakers asked to talk
for 30 min only will be less likely to drop out
[10:43] Uri Shani (IBM): maybe, skip the
domains, and play with techniques for semantic integration.
Domains can be associated as use cased to demo/example when
presenting the integration method.
[10:44] Donna Fritzsche: Uri- need to do more
work up front for that - tease out the techniques
[10:45] Judith Gelernter: If we have fewer
topics this year, we can still consider a topic next year. It's
not as though a topic will be lost.
[10:45] Mark Underwood: Cross-domain is
important for bridging human subject / human data vs. natural
science a la earth science, esp for security / privacy issues,
curation, data quality risk management etc.
[10:47] TerryLongstreth: Mark - I agree. We
would be well served with semantic integration across humanities
and Science,
[10:47] Judith Gelernter: I think I'll be
running the actual symposium at NIST this year.
[10:50] Gary Berg-Cross: The 2nd week of may
works for me, but no the first.
[10:50] Judith Gelernter: I'll check those dates
for you. 2nd week of May for the symposium at NIST
[10:50] Donna Fritzsche: @judith - what do you
think of May possibilties? off hand?
[10:50] Mark Underwood: Maybe we could ask
presenters to supply a quad chart in advance? Would that be too
Dod?
[10:50] Donna Fritzsche: what is a quad chart!
[10:51] Uri Shani (IBM): I guess i can drop from
this discussion now. Have a good day/night...
[10:51] Judith Gelernter: Scheduling depends
upon other conferences happening here. I'll have to check the
larger schedule and report back later.
[10:51] MichaelGruninger: @Judith: Can you
please check with Janet Madison about room availability in the
second week of May (or possibly even the third week)?
[10:51] Judith Gelernter: Yup
[10:51] Mark Underwood: Wikipedia "A quad chart
is a form of technical documentation used to briefly describe an
invention or other innovation through writing, illustration
and/or photographs.[1] Such documents are described as "quad"
charts because they are divided into four quadrants laid out on
a landscape perspective.[2][3][4] They are typically one-page
only; their succinctness facilitates rapid decision-making.[5]
Though shorter, quad charts often serve in a similar capacity to
white papers and the two documents are often requested alongside
one another.
[10:52] Judith Gelernter: Yup, will do.
[10:52] Donna Fritzsche: @mark - thanks
[10:52] Judith Gelernter: Let's start in
February rather than January.
[10:56] Judith Gelernter: option 1
[10:57] Mark Underwood: I had Michael's concept
in mind also
[11:03] MichaelGruninger: Another Strawman
schedule
[11:03] MichaelGruninger: January
28 Launch Session
February 4 Track X 1 11 Track Y 1
18 Track Z 1
25 Track W 1
March
3 Track U 1
10 Synthesis I 17 Track X 2
24 Track Y 2
17 Track Z 2
24 Track W 2
31Track U 2
April
7
14 Synthesis II
21 Communique I
28 Communique II
May
9 Symposium
[11:03] Mark Underwood: Gary suggested that
earlier schedule for cloud services with "lightweight" solutions
could help evolve topic depth
[11:05] Mark Underwood: Looks ok to me
[11:07] MichaelGruninger: Organizing Committee
meeting next week Dec 17
[11:08] Judith Gelernter: I can't it make that
day
[11:08] Gary Berg-Cross: I can be on a call
Thursday the 17th.
[11:08] Judith Gelernter: Okay, will do
[11:09] TerryLongstreth: I'll be there
[11:09] Donna Fritzsche: Thanks!
[11:10] Mark Underwood: I will have to move a
meeting, but I'll try to attend next week
[11:10] List of attendees: AnatolyLevenchuk,
BethDiGiulian, ChristopherSpottiswoode, Donna Fritzsche, Gary
Berg-Cross, Judith Gelernter, KenBaclawski, LeoObrst, Mark
Underwood, MichaelGruninger, TerryLongstreth, ToddSchneider, Uri
Shani (IBM), Veda Storey, anonymous, anonymous1
|
_________________________________________________________________
Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2015/
Community Wiki: http://ontolog-02.cim3.net/wiki/OntologySummit2015
Community Portal: http://ontolog-02.cim3.net/wiki/ (01)
|