Joint Ontolog-OOR Panel Discussion Session - Thu 6-August-2009    (1YV7)

Topic: Ontology Repository Research Issues    (1YV8)

Abstract: In this session, potential OpenOntologyRepository (OOR) contributors will discuss longer term issues relating to both tools and content development. See developing thoughts at: OOR/ResearchIssues    (1YV9)

Session Chair: Professor KenBaclawski (Northeastern University) - [ opening slides ]    (20A0)

Panelists:    (2066)

Archives    (206D)

Conference Call Details    (1ZKK)

Attendees    (1Z98)

Abstract    (209Q)

Resources:    (2077)

Agenda    (206I)

1. Opening by session Chair (KenBaclawski)    (206J)

2. Perspectives from the Panel    (206K)

3. Q & A and Open Discussion (All) -- please refer to process above    (206L)

4. Summary and Next Steps (KenBaclawski)    (206M)

Proceedings    (206N)

Please refer to the archives above    (206O)

IM Chat Transcript captured during the session: ...    (206P)

MikeBennett: Manufacturing standards are surely not the only kinds of
standards? e.g. financial standards    (20C4)
ArturoSanchez: Question for Michael: have you studied decidability properties
of the logical relationships among ontologies mentioned on slide 4 (e.g.,
extension). How about complexity?    (20C5)
MichaelGruninger: Response to MikeBennett: Of course, COLORE will not be
restricted to manufacturing standards, but since we are receiving our funding
through the Dept of Mechanical and Industrial Engineering our emphasis is on
manufacturing standards. In the context of the OOR project, we definitely need
to encompass all of the standards discussed at the 2009 Ontology Summit.    (20C6)
MikeBennett: Thanks Michael. I mention it because the sorts of fundamental
building blocks you mention were all needed to develop the financial industry
ontology we're working on, and I'd expect to try and align with the things you
are talking about, and with "industrial" standards generally (and UN
etc. standards e.g. FAO Country Ontology).    (20C7)
PatCassidy: MIke: do we yet have examples of 'definably interpretable' domain
ontologies interpretable with respect to some foundation ontology?  That is the
mechanism that seems to me to be the most plausible means to general semantic
interoperability, even in the absence of a repository.  It will be nice to have
examples to point to.    (20C8)
RaviSharma: Mike - mereotopology is just one example I hope? because in rela
manufacturing life, the processes of assembly and sequence of events and
processes are quite seperable, often!    (20C9)
MikeBennett: @Pat, Ravi: Mike = Michael I presume.    (20CA)
RaviSharma: Elisa - reuse is good to think but can wwe give some concrete
examples, such as a part, process, domain levels?    (20CB)
ElisaKendall: Question for Michael (when time permits): Have you looked at a
combination of ISO 11179 and 19763 as starting points for the metadata you
might need for COLORE?  I realize that they may not be complete in terms of the
kinds of characterization you might want to do, but ... the crosswalk I
mentioned might be helpful if you haven't already created something like that
yourselves    (20CC)
JoelBender: Elisa - when you talked about "Requirements range from
understanding sources ... at the ontology level to detailed provenance at the
fact/individual level" and said that was independent of the encoding - I'm
curious how you think that could actually be accomplished.    (20CD)
RaviSharma: BArt - fuzzy levels not only at probabilistic levels, but do we not
have to define probabilistIC relationships or fuzzy relationship groupings to
create or associate triples with a search or identify objects of relevance    (20CE)
ElisaKendall: Response to Ravi: There are many kinds of reuse under consideration --
ranging from general / utility / foundational (to use Michael's term)
ontologies supporting messaging (headers, channel definitions, etc. for an
event bus) and other aspects of software engineering, to reusable terms
defining missions, spacecraft, instruments ... some of which are dictated by
standards from CCSDS and others    (20CF)
BartGajderowicz: @Ravi: yes, creating probabilistic and fuzzy relationships are
key, these may be added by ontologists if they are relevant to the domain or
context, but may also be derived through machine learning based on the
experiences being represented by the ontology, and added as meta data
specifically for performing mappings    (20CG)
RaviSharma: Elisa - thanks, but we need to start thinking about how valuable
the reuse has been, almost all casses require a commonly understood
"framework", Concept or Context or many such agreements before reuse is
possible similar to code or software component or pattern reuse?    (20CH)
ElisaKendall: Response to Joel: One approach might be to have registry-level metadata
defined (standards based, possibly using ISO 11179, 19763, 1087, etc.) at the
registry level, to provide the level of provenance required per ontology, and
then incorporate annotations for the ontologies themselves, where possible.
The annotations could be managed independently of the ontologies themselves,
but don't necessarily need to be -- BioPortal does some of this, or is starting
to do some of this, but for our work, we anticipate that internal ontology and
model developers will provide this as part of the registration process.  This
is only possible if participants in the registry agree to do so, of course --
and given that we're dealing with a controlled environment for this project, we
can also provide the submission processes and requirements.  That may not be
doable on a broader, open level, though we could suggest some minimal set of
annotations that we would like to see, and even provide templates that are
representation-specific    (20CI)
JoelBender: thank you!    (20CJ)
MichaelGruninger: Response to PatCassidy: We have several papers that specify
the definable interpretations. One appeared earlier this year at the
Commonsense Reasoning Symposium (relating process ontologies) and the other
will appear in AI Journal, which relates a particular mereotopology with
classes of lattices. I just noticed that these papers are not on our website
yet -- I will post them there later.    (20CK)
MichaelGruninger: Response to Elisa: You're right -- they key is metadata about
the ontologies. I think that ISO19763 provides a good framework for capturing
the relevant ontology relationships as metadata, but we still need to
understand what relationships are essential to ontology design and reuse.    (20CL)
MichaelGruninger: Response to Ravi: Yes, of course, mereotopology is just one
example. Using application scenarios that require the integration of multiple
ontologies (as you suggest) is the right way to go in driving the development
of the repository.    (20CM)
RaviSharma: Michael - we also need to agree on what minimum type of
relationships are required for at least a minimal or upper ontology. Then Meta
data level comparisons of two or more ontology constructs (patterns like -
including metadata) can be more productive or efficient?    (20CN)
RaviSharma: Mathieu - wonderful; if there is nothing common that these exist as
independent un connected ontologies, then only benefit is by the ingeneuity of
the User as to how they use these unconnected ontologies? like going to grocery
store and coming home to cook the food?    (20CO)
MathieuDaquin: Ravi - very good point. Nothing connects two ontologies
better than their combined use/application. Hard to monitor, though.    (20CP)
RaviSharma: Todd - what do you mean by Code in repository? I thought we support
multiple languages, patterns, fraworks and domains and even globalization in
principle?    (20CQ)
MikeBennett: Terminology: in terms of metadata for ontologies, would this not
be realised by way of meta-terms for aliases, synonyms? Should synonym and a
single "alias" for reference within ODD be two separate tags?    (20CR)
ArturoSanchez: Comment for Todd and Michael: the practical issue is how can
end-users express their 'intentions' or 'intuitions' using existing ontologies
w/o having prior training in formal ontologies. Guided selection (via
interpretations using mathematical structures) might lead to choosing axioms
the end-user (the person modelling a certain situation) might not find
'intuitively' compelling. So, I think that Michael and Todd presented two
important extremes of the problem: (1) importance of formalizing relationships
among ontologies; (2) practical use of collections of ontologies. I'm
interested in the research problem of bridging these two extremes.    (20CS)
RaviSharma: Ken- your slide 8 is a representation of executable ontologies
where the data from sensors etc get filled based on the metadata values and
relationships think of this as templates and filled in forms.    (20CT)
MikeBennett: @Ken: domains is a bit of a simplification in that slide I
presume? Terms for legal, mathematical, accounting etc. have terms which
/should/ be reused in models in other domains - for instance much of business
meaning is grounded in legal realities.    (20CU)
RaviSharma: Ken- slide 11 reasoning, relationships and nature of Things
(objects) are key parameters in achieving semantics as well as
interoperability.    (20CV)
MikeBennett: @Ravi: is one issue the question of whether a given ontology
simply /asserts/ that some term is meaningful, or whether it has enough facts
defined about the thing to reliably pin down its meaning, i.e. the "context"
that was spoken of.    (20CW)
KenBaclawski: @Mike: I fully agree with you.  The diagram does not show
inter-ontology relationships.  It really must.  Is there an easy way to show
this in the diagram?    (20CX)
MikeBennett: @Ken: Possibly not, other than by creating an overall map
somehow...    (20CY)
RaviSharma: to Ken - just for record, I document that we need a map and or a
framework or an agreement on domains and some upperlevel ontology to include
others that would somehow be parts or subset of it and then there is hope for
interoperability or reuse.    (20CZ)
KenBaclawski: @Ravi: Executable ontologies is a very interesting idea.  In
practice, scientists acquire their data and then analyze it and only as the
last step do they annotate it for reuse.  It would be quite a change in
attitude to use the metadata annotations to drive the sensor data acquisition.    (20D0)
RaviSharma: Ken- The ontology part is new but for SOS earth science data we do
a metadata based query and identify the datasets of interest say using HDF_EOS
set of metadata, then we navigate to those sets, ontology would be akin to
relatime further processing these datasets for knowledge or at least
information extraction.    (20D1)
KenBaclawski: @Ravi: Is that really true?  While the ideal is to take a
top-down approach, in practice we have a large number of relatively independent
communities that deal with interoperability only after the fact.    (20D2)
RaviSharma: Ken - the community of earth scientists using satellite (NASA) data
agreed on a framework not ontology in 1996-98 and then these datasets were
generated, hence we can subset metadata values and then traverse across the
areas (geotemporal) of interest.    (20D3)
MikeBennett: There was a good talk a year or so back about a set of
relationships between ontology elements that are analogous to the owl:sameAS
but a much broader set of such e.g. sub-sets and the like.    (20D4)
RaviSharma: Mike- yes but defining a broader set of sameAs implies knowledge
built-in in the owl:sameAs    (20D5)
MichaelGruninger: Sorry, I have another call at 3:30. Great meeting today!    (20D6)
MikeBennett: @Ravi: Good point.    (20D7)
MikeBennett: @Elisa: you are describing a quality assurance framework - great
points    (20D8)
RaviSharma: Elisa - I fully appreciate your public sector need comment and
similarly for citizen services I hope.    (20D9)

Audio Recording of this Session    (206T)

   Special thanks to KurtConrad for his help in making the audio recording of this session.  =ppy    (207C)

For the record ...    (2072)

How To Join (while the session is in progress)    (1Z96)