Folks,
Thanks much to our speakers, Steve Ray, Payam Barnaghi, and Jack Hodges, and all of our participants for a very good session last Thurs, Feb. 5, 2015!
Here is the raw chat transcript for our Ontology Summit 2015 Track A “Ontology Integration in the Internet of Things” session I of Feb. 5, 2015. The session
page is: http://ontolog-02.cim3.net/wiki/ConferenceCall_2015_02_05, and the chat transcript will be added shortly.
[10:31] anonymous2 morphed into karthu
[10:35] anonymous morphed into Shadia Zaman
[11:36] anonymous morphed into Payam Barnaghi
[12:11] anonymous morphed into Yasmin Fathy
[12:17] LeoObrst: OntologySummit2015 Track A Session - Ontology Integration in the
Internet of Things - Thu 2015-02-05
[12:18] LeoObrst: Session Co-chairs: RamSriram & LeoObrst
[12:20] jack hodges: Are the slides online?
[12:20] LeoObrst: - Dr. Steve Ray (Carnegie Mellon University, USA): An Ontology-Driven
Integration Framework for Smart Communities
Describes a neutral, abstract ontology and framework that supports the vision and diverse contexts of a smart community, supporting IoT and ontology mapping
- Dr. Payam Barnaghi (University of Surrey, UK): Dynamic Semantics for the Internet of Things
Provides an overview of the use-case and requirements for semantic interoperability in the IoT with a focus on annotation, processing and information extraction
and dynamicity in the IoT environment
- Dr. Jack Hodges (Web of Things (WOT) Research Group, Siemens Berkeley Laboratory, USA): Semantic Integration Prototype for Wearable Devices in Health Care
Describes a prototype using curated biomedical ontologies to assist health care professionals in selecting appropriate wearable devices to monitor diagnosed
disorders
[12:21] LeoObrst: They will be shortly.
[12:21] jack hodges: Thx
[12:23] anonymous1 morphed into MariaPoveda
[12:26] anonymous2 morphed into CharlesVardeman
[12:27] anonymous2 morphed into Jasper Roes
[12:28] anonymous2 morphed into Ram D. Sriram
[12:29] anonymous2 morphed into Dumindu
[12:29] karthu3 morphed into karthi
[12:30] anonymous3 morphed into ChristopherSpottiswoode
[12:30] anonymous4 morphed into Bruce Nordman
[12:31] anonymous2 morphed into Mike Denny
[12:31] anonymous3 morphed into Gretchen Miller
[12:32] anonymous morphed into Donna Fritzsche
[12:32] PeterYim: Hi everyone!
[12:32] anonymous1 morphed into Johannes Hermann
[12:32] LeoObrst: Presentation material is here:
http://ontolog-02.cim3.net/wiki/ConferenceCall_2015_02_05#Presentation_Material
[12:32] anonymous morphed into Tara Athan
[12:32] anonymous1 morphed into AmandaVizedom
[12:33] anonymous morphed into Daniel Gross
[12:33] anonymous1 morphed into Dumindu
[12:33] anonymous2 morphed into Judith Gelernter
[12:35] Mark Underwood: Live tweeting at #ontologysummit - Use Tweetchat or similar
offering to join in
[12:35] anonymous1 morphed into Souhas
[12:36] Daniel Gross:
http://ontolog.cim3.net/file/work/OntologySummit2015/2015-02-05_OntologySummit2015_Ontology-Integration-In-IoT-1/
[12:38] MichaelGruninger: Presentation material at
http://ontolog-02.cim3.net/w/index.php?title=ConferenceCall_2015_02_05&oldid=15394#hid2B
[12:40] anonymous1 morphed into anonymous7
[12:45] anonymous1 morphed into Carol Hill
[12:46] Tara Athan: This "JSON Ontology" seems more like a data model than an ontology.
Am I missing something?
[12:48] Donna Fritzsche: what was the name of the company referenced?
[12:49] John Morris1: Palantir was mentioned in passing.
[12:49] Donna Fritzsche: thanks
[12:50] anonymous1 morphed into TrishWhetzel
[12:50] John Morris1 morphed into John Morris
[12:50] anonymous1 morphed into ChristiKapp
[12:51] anonymous7 morphed into Nazli Davar
[12:52] MichaelGruninger: @Tara Athan: are you referring to the lack of axioms (beyond
the implicit domain/range constraints)?
[12:54] Tara Athan: Mainly I am queasy about the term "JSON Ontology" where JSON is
a programming language, not an ontology language. So I would interpret "JSON Class" as an object type, which is not really the same thing as a "class" in the ontology sense.
[12:55] LeoObrst: @SteveRay: can you map from 1 JSON class to multiple ontology classes,
or from n JSON classes to 1 ontology class, or split/join?
[12:55] anonymous1 morphed into Daniel Puschmann
[12:56] Daniel Gross: it seems that there may be a need for some kind of meta data
that indicates how certain the meaning of a data source "data point" is.
[12:56] Daniel Gross: in particular when the data point becomes part of some analytics
processing further downstream
[12:56] Daniel Gross: that could lead to decision making models
[12:57] LeoObrst: Slide 13
[12:57] AmandaVizedom: @SteveRay's slide 9: Nice illustration of common pattern: The
"mapping" step often involves factoring out implicit/unstated aspects of the source data, necessary for correct interpretation and use, and making them *explicit* using the integrating ontology. Not only does this make the source data more usable for the intended
application, but it makes it more re-usable, now that more of its meaning is declared and machine-accessible.
[12:58] PeterYim: @SteveRay - on the work you were doing, how much of the mapping is
automated and how much of it "by hand"? (rough percentages as a reference, please)
[13:00] karthi: how is the efficiency of ontology mapping process in real time systems?
how fast is it?
[13:00] AmandaVizedom: <there is a funny clopping noise; it sort of sounds as if @SteveRay
is riding a horse. 
>
[13:01] Mark Underwood: FYI Steve Rays' slide 16 all abt standards - FYI - Our track
D is focused on standards issues
[13:01] JackRing: Where do you note the existence of known bugs in the devices?
[13:01] Tara Athan: There is a flavor of OBDA (Ontology-based Data Access) here, where
the JSON is essentially a dump from a OO-database. But in OBDA, one is usually careful to not confuse object types with ontology classes. Instead there is an explicit "Mapping" step to relate them. It may happen that there is a 1-1 correspondence between object
types and ontology classes, but they should not be consider to be the same thing.
[13:03] Donna Fritzsche: great point - mapping to neutral ontology so that interoperability
can be built in
[13:03] JackRing: the clopping noise is someone keyboarding not muted.
[13:03] RaviSharma: @Steve: Palantir also use ontologies for Enterprise wide Data Dictionaries
and services and are present in their consulting in Sacramento rea I believe.
[13:04] RaviSharma: @steve. As we discussed we need ontology of standards?
[13:04] Mark Underwood: @Donna: Maybe a neutral ontology "vetted" through repeated
Github deployments
[13:04] Gary Berg-Cross: @steve Since the topic is ontology integration, a Q is, did
you leverage the Semantic Sensor Network Ontology and if not why not?
[13:04] Donna Fritzsche: interesing
[13:05] SteveRay: @Leo: Yes, you can map one to many or many to one, as on slide 9.
[13:06] RaviSharma: @Steve. Another thought is create integration patterns for sensor
types that are more like web services standards so the execution and sharing are proven patterns?
[13:06] anonymous1 morphed into Yasmin Fathy
[13:07] SteveRay: @Peter: Basically all of the mapping is done by hand. I have toyed
with the idea of using something like WordNet to suggest mappings based solely on the terms used, but clearly that will be imperfect. So the tool automates all the drudgery, but leaves the recognition of concepts to the human.
[13:07] AnatolyLevenchuk: ISO 15926 was developed as neutral engineering ontology in
a situation of "standard quagmire". Mapping to the neutral ontology as a central selling point. Still not used much.
[13:08] SteveRay: @Karthi: The process runs very quickly, basically in one second or
so. Haven't run any tests with huge data sets, though.
[13:08] karthi: @steveRay, What kind of ontology matcher is used?
[13:09] SteveRay: @Jack: Could you expand on your question about bugs?
[13:09] JackRing: PB #5 should include Errors
[13:10] RaviSharma: @Payam
[13:10] SteveRay: @Tara: I agree with you. I probably shouldn't call the JSON structures
an ontology. Regarding the correspondence between object types and classes, I agree they aren't the same thing, and the mapping stage handles that.
[13:11] SteveRay: @Ravi: An ontology of standards would be one more meta-level up.
I'm still working on the contents inside each standard.
[13:12] Mark Underwood: @SteveRay - Which of the smartgrid standards are easiest to
work with from an ontologists' perspective?
[13:13] JackRing: PB#10 does not indicate probable error in the data
[13:13] Donna Fritzsche: @anatoly - thanks reference:
http://en.wikipedia.org/wiki/ISO_15926
[13:13] SteveRay: @Gary: I looked at the SSN ontology, but it focuses more on the relationship
between phenomena and the effect they have on sensors, leading eventually to data. That is more upstream from the interoperability of the data that we were focused on.
[13:15] SteveRay: @Karthi: Regarding ontology matcher, are you talking about some sort
of automated system that can recognize similar patterns? We don't have anything like that. I believe Natasha Noy was considering that when she was at Stanford.
[13:17] karthi: @SteveRay, yes I am meaning about automated system, where a new data
can automatically align itself with the existing base/neutral ontology
[13:17] AmandaVizedom: @SteveRay [13:13] - That's interesting. Do you have a lot of
analysis going on that combines data sets? I ask because in my current work, there is a need to have that info included, in order to support combining data across data sets and Big Data - type analyses.
[13:18] MikeBennett: Recognizing similar patterns may or may not be a reliable way
of recognizing corresponding meanings. Depends on the granularity. I would recommend some use of grounding of concepts across different ontologies.
[13:19] SteveRay: @Mark: Well, ASHRAE SPC201 (sometimes called the FSGIM) has been
converted from UML to OWL (by me), so it is in good shape. I have also converted EMIX, OpenADR, Energy Interoperation (all OASIS standards) with reasonable success. Generally, the PROTOCOL standards are less amenable, because they typically conflate the model
with the message syntax. The ASHRAE SPC201 standard is explicitly an information model and not a protocl, hence it's easier to work with.
[13:19] RaviSharma: @Payam - we need to have or use if exist - domain based sensor
standards healthcare as we will hear in next talk, grid like Steve mentioned, agri-climate sensors data, etc so that ontologies can address subsets at a time. We today deal with parsing out relevant traffic by IP and other filters, for IPv6 and beyond, we
need to think of parsing out relevant IOT by domain, then only apply ontology for integration with other info and reach target outputs, whether energy billing or diagnostics. If we try to mix many domains in tegration, we will run into semantics across domains
and perhaps lead to confusion? would like to hear your comments. Namely integration by domain.
[13:20] Mark Underwood: @SteveRay - thx, will try to mention in our last Track D Session
[13:21] SteveRay: @Mark: One standard that has been cripplingly difficult so far is
BacNET - a protocol standard.
[13:22] SteveRay: @Amanda: We could talk about trying out our framework with some of
your data sources if you like.
[13:25] Gary Berg-Cross: @ Steve If you found the SSNO too upstream did you align
with any other existing ontology?
[13:26] AmandaVizedom: @Steve, my comment was more directed in understanding what looks
like a use case difference. You said that you didn't find that you need to model the kind of info the SSNO is focused on. I'm interested in how that might relate to the use case. Or more precisely, on whether the difference between your use case and mine might
reflect a pattern in what's needed for certain application scenarios.
[13:26] RaviSharma: @Payam - slide 25 is similar to what was done by NASA in earth
science but only addressing their repositories, but subset and metadata based searches on IOT are more complex as you can not apply same metadata for different THINGS?
[13:26] AnatolyLevenchuk: ISO 15926 is 4D extensional ontology, now I know that this
more convenient to be called "dynamic semantics". It was developed to describe equipment change in an industrial plant repair or redesign. E.g. device change in IoT particular configuration.
[13:27] John Morris: Naive question -> "semantics disappears" at higher layers? OK
for performance, but is it strictly true that "semantics disappears"? There are still semantics, just not embedded in a very heavy software/modeling framework?
[13:27] JackRing: We have not considered the IoT when a million of the Things are software
bugs thus erroneous data.
[13:28] MikeBennett: I think there are multiple uses of the word "semantics" in play,
e.g. Semantic Web semantics (consistent patterns, deductive closure semantics) versus meaning more broadly in a network of concepts (semantics in terms of relationships between concepts, with some grounding of concept). It sounds to me that both have a role.
But calling something "not semantics" may be a source of confusion IMHO.
[13:32] karthi: how to validate the semantics created?
[13:33] SteveRay: Gary: I aligned with schema.org, qudt.org, W3C's geo ontology, Dublin
Core, and informally some life-logging ontology concepts.
[13:34] LeoObrst: @PayamBarnaghi: what about mal-models, i.e., intervening models that
are deliberately false or distorted, to interrupt an interpretation?
[13:37] RaviSharma: @Payam - agree with Leo, great talk.
[13:38] anonymous1 morphed into Kerstin Forsberg
[13:39] Payam Barnaghi: @RaviSharma: that's correct you can not apply same metadata
for different Things; but if they have some common attributes like location, time and type then on the index/discovery layer you only use them; so your indexes are interoperable but your models could even be different
[13:41] Payam Barnaghi: RaviSharma:mix many domains in tegration-> agree- but in some
cases we may be able to use a common base; like SSN for sensors but for health sensors you may add a new module with more medical sensor related attributes; but the core attributes are interoperable
[13:42] Payam Barnaghi: @Gary Berg-Cross--> SSN is an upper-level ontology it is aligned
with DUL
[13:42] RaviSharma: @Jack - are all wearable types having different API types or is
there a standard so integration can align information elements that supplement each other.
[13:44] Payam Barnaghi: @karthi: this is a simple example: iot3.ee.surrey.ac.uk/SSNValidation/
[13:44] RaviSharma: there is echo / feedback and jitter so please request presenters
to turn off speaker
[13:45] MariaPoveda: @Payam how do you generate the tag cloud in iot3.ee.surrey.ac.uk/SSNValidation/?
[13:45] Payam Barnaghi: @John Morris: I meant no RDF at higher layers; but yes the
notion of semantic still be thre
[13:46] AmandaVizedom: @PayamBarnaghi: re [13:39][13:41]: Do you see this as signficantly
different between #IoT case and a more general case of combining and/or reusing data from heterogeneous sources? I see many similarities between this and, e.g., researcher re-use of accumulated multi-source data (from varied sensors, programs, analyses, etc.).
[13:47] Payam Barnaghi: @LeoObrst: about mal-models--> that's another problem; validators
and compliance tools can probably help
[13:47] MariaPoveda: @Karthi there is a similar service to the one Payam point out
but for ontologies in any domain in http://oops.linkeddata.es/ It checks
several modelling decision, not really the semantics as it only takes the ontology as input, not the requirements.
[13:48] AmandaVizedom: @PayamBarnaghi: continued from my [13:46] -- noting, of course,
that in the #IOT case there may be a greater real-time / velocity factor.
[13:48] Payam Barnaghi: @MariaPoveda: it is based on the most common terms that we
collect from the submitted ontologies;
[13:49] anonymous1 morphed into BobbinTeegarden
[13:49] MariaPoveda: @Payam I mean whether you have any service or api providing it,
I would like to reuse it
[13:49] RaviSharma: @Jack - two way communication and passive data collection - would
complexities in handling target info differ in underlying ontologies
[13:49] Gary Berg-Cross: @Payam Yes i would think that people could use SSN the way
that Jack is describing even if it is only a part of it,
[13:50] Payam Barnaghi: @JackRing--> SSN is an upper-level ontology so doesn't include
all the attributes but has component for expressing quality related attributes
[13:51] John Morris: Here's my question, when it's possible. Healthcare and wearables
are also in the domain of "alarm fatigue", i.e. the well-documented but serious issue of too much data delivered to front-line healthcare staff. Modeling alarm semantics is probably an important way of meeting this challenge -- even a full alarm domain ontology.
I'm wondering if the Siemens team is addressing this issue? It seemed like a relevant question given the subject of the presentation.
[13:52] Payam Barnaghi: @Gary Berg-Cross: yes, exactly; SSN was designed with the assumption
that people will extend it to use the ontology
[13:54] Mark Underwood: @JackHodges Can this work be scaled up for decision support
with eg the Siemens Somatom CT scanners?
[13:54] Gary Berg-Cross: @payam In track B last week we tried to talk about some of
these SSN extensions and uses (your work was cited) and that is one reason that I raise it here for connection.
[13:55] AmandaVizedom: @JackHodges - re slide 10 - Can you give a more particular sense
of what you mean by "large"?
[13:56] Gary Berg-Cross: I think that potential hooks in an ontology is a good strategy.
One may illustrate how to connect with other things outside of an ontology module or pattern.
[13:57] Payam Barnaghi: @Gary Berg-Cross: unfortunately I couldn't attend track B session
last week
[13:59] Gary Berg-Cross: @Jack You mentioned "Some have substantial structure representing
content." Did this include modular approaches (SSN is modular) and if so what difference did it make for you? It would seem that with "large" ones a module makes focused used easier. Did it?
[13:59] LeoObrst: @JackHodges and all: what if you want to push/update data to a particular
IoT thing? E.g., push a shock or increase a dose on a medical device?
[13:59] RaviSharma: @jACK - Do we think that IoT limit to data level and aggregation
with other ontologies and decision support be done elsewhere namely in applications and reference docs such as Snomed etc. to arrive at resulting data communication back to IOT, would that make the models simpler but might delay action? unless devices are
intelligent to do processing and only communicate the actions or reports. I am thinking of it in context of 5 ontologies used by you in example.
[14:00] Tara Athan: @SteveRay - if SSN seems to upstream for you, you might like OGC's
O&M better (http://www.opengeospatial.org/standards/om).
However, there are some known "issues" with O&M and the rate of upgrade of this model is slow since it is already used in production. There was an earlier ontolog-forum talk about this.
[14:00] SteveRay: @Tara: Thanks. I'll check it out.
[14:01] JoelBender: @TaraAthan: 404 error on that URL
[14:01] TerryLongstreth: @JohnMorris: alarm fatigue might be more germane to the device
wearers
[14:01] JoelBender: @TaraAthan: oops - never mind
[14:01] PeterYim: try not to post url's with punctuations "." ")" etc.
[14:02] SteveRay: @Joel: I got an error as well. Did it work for you eventually?
[14:02] JoelBender: <http://www.opengeospatial.org/standards/om>
[14:02] John Morris: Thank you.
[14:04] MariaPoveda: <I'm sorry I have to leave in 5>
[14:05] RaviSharma: can u hear
[14:05] Payam Barnaghi: @AmandaVizedom: there are lots of similarities between semantics
on the web and IoT; but there are also significance differences; especially velocity and dynamcity are different; Iot is often streaming data with errors, variable qualities, ... and network and energy cost of data collection and publication are also important
[14:05] RaviSharma: ok
[14:05] JackRing1: @Payam Barnaghi the "upper level" of any sensor is a) data and b)
probable error. But I am not concerned with the mechanics of the ontology as much as about the naivety of the model.
[14:05] John Morris: @TerryLongstreth - yes, you are correct, alarm fatigue is relevant
to the device wearers certainly. It's also (as documented by research at Johns Hopkins for example) an on-going problem, not solved. The efforts to solve that have been successful seem to be going into a direction that this community could contribute to.
[14:06] PeterYim: @JackHodges - you mentioned something about the lack of references
in the ontologies, but for DOID, for example, they claim to have very rick external references - ref.
http://do-wiki.nubic.northwestern.edu/do-wiki/index.php/History
[14:07] PeterYim: ^ ... you did pick very solid ontologies!
[14:08] TerryLongstreth: @JohnMorris +1
[14:08] RaviSharma: thanks
[14:08] Payam Barnaghi: @JackRing1: some of the work is now continued at:
http://www.w3.org/2015/spatial/charter
[14:08] Daniel Gross: Thank you
[14:08] PeterYim: ^^ ref. your slide 6
[14:10] SteveRay: @Tara: Do you know how the opengeospatial.org ontology relates to
the SSN ontology?
[14:12] Tara Athan: @SteveRay - If you mean a formal alignment, I am not aware of one.
[14:12] jack hodges: We are working on alarms in terms of how activities are subscribed
to, so the HCP using our platform would define alarms for a particular patient and set of properties and then be advised accordingly.
[14:12] Gary Berg-Cross: Jack mentioned "Bridge ontologies". Did either of the other
speakers use or create something like this and what is their opinion on the role of Bridge ontologies?
[14:13] PeterYim: @JackHodges - further to my earlier (verbal) comment, some references
may be available from the BioPortal ontology repository - http://bioportal.bioontology.org/
[14:13] Mark Underwood: @Gary - I wondered the same
[14:13] AmandaVizedom: @PeterYim, could you repeat here what your question was regarding
availability of ontologies via URIs?
[14:13] MikeBennett: Someone is participating in another call
[14:14] SteveRay: @Gary: I suppose you could think of our base ontology as a bridge
ontology.
[14:15] PeterYim: @[xx:13] AmandaVizedom - my question was captured at [xx:06] (...
I might have heard Jack wrong, though)
[14:15] AmandaVizedom: @Peter - Oh, I see. Thanks.
[14:16] Gary Berg-Cross: @Steve I guess I thought of your base ontology as the skeleton
for the content (Resource for example) and a bridge ontology as something to connect selected parts of ontology being reused.
[14:17] Dumindu : @ jack Could the model you presented be integrated with image processing
where it could use more effectively in industrial applications ? even for biomedical applications with image processing necessities ?
[14:17] TerryLongstreth: "The quality of data is not plain"
[14:17] PeterYim: ^^ [11:06] PeterYim: errata: should read: "... have very rich external
references"
[14:18] jack hodges: @leo we are working on activities/actions that could be sent to
devices, but I am more interested in devices responding to events but, through the use of ontologies, understanding what those events (and from what sensors/devices) mean so that they can respond appropriately.
[14:19] SteveRay: @Gary: Hmm. The base ontology does contain associations among its
concepts. If you are talking about connections to the data sources, that is what the mappings are doing. Am I missing something?
[14:19] Donna Fritzsche: To state a different way: Need to capture and model data provenance
and reliability of sources when building out truth models.
[14:20] MikeBennett: @Donna +1. "Meaning is not Truth". Ontologies such as Prov-O,
and a good ontology of known data quality metrics would be of value.
[14:21] Donna Fritzsche: did not mean truth literally! - quality of data for the context
is corect
[14:21] AmandaVizedom: @PeterYim @JackHodges - FWIW, out of curiousity I did a quick-check
on DOID and SNOMED, since both were mentioned, and both are indeed on bioportal with dereferencable uris per concept (see
http://bioportal.bioontology.org/ontologies)
[14:22] PeterYim: @SteveRay: your "clopping" sound is probably an audio feedback that
is suppressed (but not well enough) by your equipment ... you might be able to get rid of it if you use a headset rather than an open mic and an open speaker (which I assume you are using.)
[14:23] SteveRay: @Peter: Thanks. I should do some testing sometime. I don't have any
"clopping" during Webex or GotoMeeting sessions, so I'm told, but maybe this bridge is more demanding (or something).
[14:23] PeterYim: @[11:21] AmandaVizedom - see
http://bioportal.bioontology.org/ontologies ( ... pls drop
the ")' to resolve url properly)
[14:24] RaviSharma: thank Leo
[14:24] PeterYim: great season!
[14:24] LeoObrst: Ontology Summit 2015 Track C (Decision Making in Different Domains
): Thu 12-Feb-2015
[14:24] LeoObrst: Ontology Summit 2015 Track C (Decision Making in Different Domains
): Thu 12-Feb-2015
[14:24] Mark Underwood: Excellent presentations - thanks to presenters & organizers!
[14:25] SteveRay: Thanks for the privilege of presenting.
[14:25] MarcelaVegetti: Thanks all. Very interesting session today!
[14:25] Donna Fritzsche: thanks!
[14:25] MarcelaVegetti: bye
[14:25] TerryLongstreth: Data quality is a very complicated subject; I may want to
be collecting examples of false information
[14:25] Payam Barnaghi: Thank you. bye
[14:25] LeoObrst: Thanks, All!
[14:25] Gretchen Miller: Thanks!
[14:25] Mark Underwood: @TerryLongstreth - Did this come up in last year's big data
summit?
Thanks!
Leo and Ram
Track A Co-champions
_______________________________________________
Dr. Leo Obrst The MITRE Corporation, Information Semantics
lobrst@xxxxxxxxx Cognitive Science & Artificial Intelligence, CCG
Voice: 703-983-6770 7515 Colshire Drive, M/S H317
Fax: 703-983-1379 McLean, VA 22102-7508, USA
