Bobbin
Very interesting...
More later..
On Feb 1, 2012 5:07 PM, "Bobbin Teegarden" < teegs@xxxxxxxxxxxxx> wrote:
We keep dancing around what I would call process (rules/algorithms are sort of 'micro-processes'), and it would be great to hear thoughts about an eXecutable ontology -- one that combines in one structure, data with the processes that use them; that combine the 'what' with the 'how'; that might combine the wave view (process) with the particle view (graph of structure). Would that bring us to a tTOE -- a technical TOE (Theory of Everything as in physics), and is it logically possible.
I know several folk (including myself) in the Summit crew who have thoughts on how to include activities/process in ontologies, and how possibly execute them.
So what Track would these fit in, and is there interest?
And by the way, rules do the heavy-lifting in reasoning over ontologies and their instances. But rules (axioms, inference rules, consequence relations) are just as declarative as ontologies, i.e., ontologies expressed in a logic, thereby being a “logical theory”, use the declarative apparatus of logic.
So the transformation I spoke of are transformations from one declarative representation to another (mostly or entirely) declarative representation. Otherwise, humans do all the programming imperatively, looking only at the ontologies, conceptual models, conceptual schemas as “guidance”. And from that process, there are many algorithms that could be created, perhaps infinitely many, depending on the space/complexity of the computation.
I hope this helps to clarify things.
Thanks,
Leo
From: Obrst, Leo J.
Sent: Wednesday, February 01, 2012 6:27 PM To: Ontology Summit 2012 discussion Subject: RE: [ontology-summit] Ontologies are not algorithms [was: Ontolgizing rain & snow]
Jack,
Algorithms are ways to do things. How.
Ontologies, as a subclass of declarative methods, are ways to describe things. What.
From an ontology, one can spawn multiple algorithms that use the ontology, but this requires a set of transformations from the declarative to the imperative. And it will be multiple, because for any statement of what to do there are many ways how to do it.
If you are working in a declarative paradigm, then the way to do it is is to apply declarative transformations as far as you can, to get close to a runtime how that preserves the what as much as possible.
E.g., from a first-order ontology, you use knowledge compilation techniques (examples: reduce to Horn Logic approximations, use implicants/implicatures, etc.; there are a range of tools available, but this is another partial research thread). Or you simply use a first-order reasoner that uses the ontology directly; however, FOL reasoners will be slower than other reasoners, because they deal with more expressive logical expressions.
Logic programming (Prolog, Answer Set Programming), by circumscribing the FOL syntax, along with potentially using some non-declarative constructs (the “cut” operator, ordering restrictions, a form of negation that is not quite logical, i.e., negation by finite failure), etc., can closely preserve the ontology and approximate its declarative entailments, etc., in more efficient runtime reasoning.
But if you want to transform an ontology to an imperative algorithm: good luck. Mostly these transformations are done by human programmers. One can talk about semantics-preserving programs, etc., but it’s pretty much ad hoc-land. One can annotate programs (like “documentation”) as to what the imperative constructs are supposed to mean, but it is really a very loose correlation.
That’s the difference.
Thanks,
Leo
Dear Nicola, I am aware of several kinds of computational ontologies. I was interested in a specific example of what you had in mind when declaring that an ontology is not equivalent to an algorithm.
Ontologies that I have seen express semantic equivalences and other relationships. Some are even loaded with first order predicates that look very much like spaghetti code (which I thought was banned in the 1980's).
However, I am not interested in arguing the point. The simple fact is that systems and systems engineering need semantic transformers. If ontologies don't do semantic transforms (e.g. AP233) then there may be some residual utility in ontologies in Big Systems but not the degree I had hoped.
ps. 'algorithm' predates computer science by several centuries.
On Feb 1, 2012, at 2:57 PM, Nicola Guarino wrote:
Dear Jack,
sorry for this late answer to your request. There are various definitions of (computational) ontologies adopted by the community, it shouldn't be difficult for you to find them, as well as actual examples of ontologies.
Speaking for myself, I spent a good fraction of my research career trying to clarify, in a rigorous way, what a computational ontology is (see for instance the paper by Daniel Oberle and myself on the Handbook of Ontologies, 2nd edition). Basically, I agree with the most popular definition "An ontology is a specification of a conceptualisation", which requires however a careful clarification of what a conceptualisation is (but this is a long story, although discussed in various papers). In logical terms, an ontology is just a logical theory expressing a set of meaning postulates.
In my opinion, none of the current definitions of a computational ontology is compatible with the idea that ontologies are algorithms. If you want to use classing terms of computer science, you can perhaps compare an ontology to a data structure, but not to an algorithm.
In any way, this is not a topic to be addressed in this summit. So, if people are still interested, I would suggest them to move the discussion to the general Ontolog list.
On 31 Jan 2012, at 04:14, Jack Ring wrote:
Gee, I made a very explicit statement in my brief last week ---- that ontology is algorithm --- and you and Nicola immediately and clearly disagreed. I followed up with a request foer an example; of an ontology so we could get down to specifics but have not yet received one. Meanwhile an example of Cyc ontology appeared and I gave some specifics from a systemist viewpoint.
If you think that emergence, a very important phenomenon in systems, and system models is not a core issue for ontologists to deal with then I guess the intersection of SE and ontology in this forum may turn out to be only the transform from drawing (pump 102) to purchase order (sku2058), i.e., engineering of systems.
On Jan 30, 2012, at 6:18 PM, Obrst, Leo J. wrote: Jack, it's not that this is not interesting, and I get the metaphor, but everyone needs to make the metaphor explicit. Otherwise, it seems we are veering all over the place and not focusing on our themes for THIS Summit. The next summit or a future summit can address human-nature co-systems, and even this summit can, where it is appropriate to our focus.
Make it explicit what the contribution is, focus. We simply don't have the open-endedness of the whole Ontolog Forum, where nearly anything and everything goes.
Thanks,
Leo
-----Original Message-----
From: ontology-summit-bounces@xxxxxxxxxxxxxxxx [mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Jack Ring
Sent: Monday, January 30, 2012 7:29 PM
To: Ontology Summit 2012 discussion Subject: Re: [ontology-summit] Ontolgizing rain & snow [was: Track 1&2 Joint Mission and Session Abstracts]
Leo,
Apologies for the ambiguities. The subject is not rain and snow. Those are the metaphors. The subject is emergence and the special confusions of triple-point systems.
OBTW, perhaps ecologies are systems only in the minds of humans. if you don't want to take up natural vs. human-imputed systems this year so be it. However it is already a big confusion in the societal demand for systems. For example sponsors do not know that more than 90% or Mother Nature's experiments fail.
On Jan 30, 2012, at 5:20 PM, Obrst, Leo J. wrote:
Sure, human engineered systems also contain humans. That's what systems engineering is all about. And teleological arguments do show correspondences to human-engineered systems (function of the heart in the human body). But remember we are addressing human-engineered systems. Ecologies are systems and are wider than human-engineered systems, but obviously also affect the latter. Perhaps ecologies, especially human-influenced ecologies are super-systems, and we should address these here, but I think we are veering off.
I suggest just like we are not addressing natural ecologies, at least not in this summit. Or at the least: not in this thread yet. We are ontologizing rain and snow.
Thanks,
Leo
-----Original Message-----
From: ontology-summit-bounces@xxxxxxxxxxxxxxxx [mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Jack Ring
Sent: Monday, January 30, 2012 7:01 PM
To: Ontology Summit 2012 discussion
Subject: Re: [ontology-summit] Ontolgizing rain & snow [was: Track 1&2 Joint Mission and Session Abstracts]
I think you will have a difficult time explaining why natural systems are not human-presumed systems. And closer to reality, yet, if human engineered systems contain N humans as active components then what?
On Jan 30, 2012, at 4:55 PM, Obrst, Leo J. wrote:
Folks, this might be a discussion thread more appropriate for the more general [ontolog-forum], since it doesn't really address human engineered systems, but instead natural systems.
What do you think?
Thanks,
Leo
-----Original Message-----
From: ontology-summit-bounces@xxxxxxxxxxxxxxxx [mailto:ontology-summit-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Jack Ring
Sent: Monday, January 30, 2012 6:24 PM
To: Ontology Summit 2012 discussion
Subject: Re: [ontology-summit] Ontolgizing rain & snow [was: Track 1&2 Joint Mission and Session Abstracts]
Doug, Thank you for this. As is probably obvious by now or will be by the end of this message I am not a practicing ontologist.
I am struck by several presumptions and gaps in the Cyc example. Perhaps I just can't read right or perhaps these are not proper ontology stuff. As I said in the last track, I am not sure I am looking for ontology as the means for systemist interoperability. Perhaps I seek some yet-to-be-conceived semiotic transformer. At any rate, humor me regarding the following:
The example doesn't say where rain comes from. Mentions clouds but clouds are not rain or they wouldn't be "up there."
I am looking for the notion of emergence. Rain happens after raindrops occur. Why do raindrops occur? And why snowflakes rather than raindrops? And why not mention fog as well? And Relative Humidity?
OBTW, rain water is not fresh. Raindrops condense on particles, every raindrop has one therefor rainwater is laden with particles (especially ones resulting from cloud seeding).
OBTW, Raindrops fall on me in Arizona while the sun is shining. No storm evident. Just an occlusion of a low and high pressure trough waaay up there.
I am not trying to be smart alecky here. It is just that systems work challenges you to think beyond the active entities and consider the not's as well. Otherwise Unintended Consequences are born.
I maliciously mentioned rain and snow because H2O has a triple point, vapor, liquid, solid depending on pressure, temperature, etc., Not all substances have such triple point. Methinks "SYSTEM" does which is the root of much confusion therefore a challenge for ontologists.
Thanks for your attention.
Jack
On Jan 26, 2012, at 10:58 AM, doug foxvog wrote:
Matthew West wrote:
I understand your view. How shall you handle rain and snow?
Rain & snow refer to physical precipitation particles, the precipitation in
bulk, the process that produces the precipitation, storms as events, and
storms as objects.
Cyc's representation of these different, but related things (leaving out
comments and some additional statements) includes:
(isa PrecipitationParticle ExistingObjectType)
(genls PrecipitationParticle Particle)
(genls PrecipitationParticle InanimateObject-Natural)
(isa RainProcess ProcessType)
(genls RainProcess PrecipitationProcess)
(isa SnowProcess ProcessType)
(genls SnowProcess PrecipitationProcess)
(isa Rainwater ExistingStuffType)
(genls Rainwater (LiquidFn Water-Fresh))
(isa SnowMob ExistingStuffType)
(genls SnowMob (SolidFn Water))
(isa Snowflake ExistingObjectType)
(genls Snowflake PrecipitationParticle)
(genls (MobFn Snowflake) SnowMob)
(isa Raindrop ExistingObjectType)
(genls Raindrop PrecipitationParticle)
(genls Raindrop Rainwater)
(relationAllExists outputsGenerated PrecipitationProcess
(MobFn PrecipitationParticle))
(relationAllExists outputsGenerated RainProcess (MobFn Raindrop))
(relationAllExists outputsGenerated SnowProcess (MobFn Snowflake))
(isa StormAsObject ExistingObjectType)
(genls StormAsObject InanimateObject-Natural)
(relationAllExists physicalParts StormAsObject CloudInSky)
(isa RainStormAsObject ExistingObjectType)
(genls RainStormAsObject StormAsObject)
(relationAllExists physicalParts StormAsObject CloudInSky)
(relationAllExists physicalParts StormAsObject (MobFn Raindrop))
(isa SnowStormAsObject ExistingObjectType)
(genls SnowStormAsObject StormAsObject)
(relationAllExists physicalParts StormAsObject CloudInSky)
(relationAllExists physicalParts SnowStormAsObject (MobFn Snowflake))
(not (relationExistsAll doneBy PrecipitationProcess StormAsObject))
(comment
(not (relationExistsAll doneBy PrecipitationProcess StormAsObject))
"A StormAsObject would include Duststorms, which don't (necessarily)
include precipitation.")
(relationExistsAll doneBy RainProcess RainStormAsObject)
(relationExistsAll doneBy SnowProcess SnowStormAsObject)
(isa StormAsEvent ExistingObjectType)
(genls StormAsEvent ImmediateWeatherProcess)
(isa RainStormAsEvent ExistingObjectType)
(genls RainStormAsEvent StormAsEvent)
(relationAllExists subprocesses RainStormAsEvent RainProcess)
(relationAllExists doneBy RainStormAsEvent RainStormAsObject)
(isa SnowStormAsEvent ExistingObjectType)
(genls SnowStormAsEvent StormAsEvent)
(relationAllExists subprocesses SnowStormAsEvent SnowProcess)
(relationAllExists doneBy SnowStormAsEvent SnowStormAsObject)
(isa SnowStormAsObject ExistingObjectType)
(genls SnowStormAsObject StormAsObject)
(relationAllExists physicalParts SnowStormAsObject (MobFn Snowflake))
Jack
On Jan 26, 2012, at 6:09 AM, Matthew West wrote:
The main problem here is one of different people using terms
differently. Hardly an ontological problem per se, but certainly a
problem that causes confusion in developing ontologies.
This is always a problem for ontologists. The different meanings have
to be teased apart.
Interestingly as a 4 dimensionalist I don't recognise endurants at all,
but I do recognise activities, physical objects, and participants. Under
this world view all individuals (including activities, physical objects
and participants) are spatiotemporal extents, and you discover that an
activity consists of its participants, where a participant is the state
of a physical object that participates in some activity. So I recognise
the things you talk about. However, I would assign the term "system" to
the physical object the participant is a state of.
I would not restrict the term "system" merely to physical objects. But
having multiple clearly defined concepts which different people use that word
for in different contexts, is fine. They just need different URIs.
-- doug
Regards
Matthew West
Information Junction
Tel: +44 1489 880185
Mobile: +44 750 3385279
_________________________________________________________________
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/OntologySummit2012/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2012
Community Portal: http://ontolog.cim3.net/wiki/
...
_________________________________________________________________
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/OntologySummit2012/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2012
Community Portal: http://ontolog.cim3.net/wiki/ (01)
|