W3C home > Mailing lists > Public > public-xg-ssn@w3.org > June 2009

RE: [ExternalEmail] Re: another relevant bit of ontology

From: <Kerry.Taylor@csiro.au>
Date: Thu, 4 Jun 2009 22:46:15 +1000
To: <public-xg-ssn@w3.org>
Message-ID: <B484B32EAEABE14AA5409575229CECF3A66072C50A@EXNSW-MBX05.nexus.csiro.au>
All,

I firmly believe that we should limit our language in the ontology to OWL 2 DL, and also to attempt to
use as much of OWL 2 DL as we sensibly can.

Although many people consider that "semantics" arises simply due to the
consistent naming of things, I don't see this gets us much ahead -we can do that in any
syntax we like -- and there are many to choose from! The really good thing about OWL, though, is its computational
interpretation of the relationships amongst ontological terms  -- that is domain-independent reasoning
over the domain of discourse (of course other logical formalisms do this too).

As for use cases -- by all means any kind of reasoning should be anticipated, provided that it
does not disrespect the OWL 2 SROIQ reasoning. We certainly should not be thinking that our ontology can only be used
for something, or in some way, but not for something else!

But then, to answer John's question : "how to constrain what kinds of reasoning  we mean to support?" We don't need
to constrain anything- but then (in my view) neither should we attempt to define any rules to support rule-based reasoning, nor
a thesaurus to support SKOS-like reasoning, nor abductive reasoning, nor..... On the other hand,
RDF comes for free, and our use cases should reflect that.

Kerry

________________________________
From: public-xg-ssn-request@w3.org [mailto:public-xg-ssn-request@w3.org] On Behalf Of Michael Compton
Sent: Thursday, 28 May 2009 11:22 AM
To: John Graybeal
Cc: Cory Henson; public-xg-ssn@w3.org
Subject: [ExternalEmail] Re: another relevant bit of ontology

I think I'm in agreement with both Amit and John in that I think semantics is (here comes a gross simplification) really about description and relationships, and that you should reason about those things (and the data that they describe) by whatever method works best.  No need to stop with OWL, rules and relations - though, I think you can get a long way with that much.

If you are concerned with subsumption, satisfaction etc, then OWL/DL reasoning is the way to go, but if your semantics tells you that to work out that it's winter you actually need observations from sensors of a particular type from a particular area and a particular sort of temperature/temporal/atmospheric model written in C, then use the OWL to compute the sensors of the right type, and your spatial reasoning to find the ones in the right area and then use the model to determine if it's winter.  If the process tells you that you need to assert new triples that might be used by some other sort of reasoning, go ahead - but you'd better get the logic right.

I guess I'm just saying the obvious thing that semantics is supposed to enable all these things and not be constraining.  So yes why not "any reasoning over any information associated with sensors".

Michael




On 27/05/2009, at 5:20 , John Graybeal wrote:

OK, I think that one explicitly addressed the fact they were drawing inferences about the reasons the data has problems.  For example, "In this section, we explore how our reasoning system can evaluate these data to signal the plausibility of several hypotheses to explain their correlations." (p.7)

I assume inference over observational data fits into the scope of the project... In my naive world, I assume any reasoning over any information associated with sensors fits into the project.  If the project is to semantically describe sensors, how would you constrain what kinds of reasoning we mean to support?

John


On May 26, 2009, at 7:55 AM, Cory Henson wrote:

Sorry for confusion, I was referring to the paper you sent to start this thread.

[1] http://efg.cs.umb.edu/pubs/SensorDataReasoning.pdf

-Cory


On Tue, May 26, 2009 at 10:50 AM, John Graybeal <graybeal@mbari.org<mailto:graybeal@mbari.org>> wrote:
Cory,

Please clarify which link you mean by "this paper"?  Thanks.

John


On May 26, 2009, at 7:03 AM, Cory Henson wrote:

The paper seems to be discussing deductive rule-based reasoning with sensor data. While this is definitely useful for certain tasks (i.e., checking constraints and data validity) it may be insufficient for other purposes such as interpretation and explanation of sensor data. For example, the rule on page 5 says that if the temperature is less than zero Celsius then it must be winter.  But, is it winter because it is cold, or is it cold because it is winter? Or is it cold because the sensor is in a freezer? Interpretation of observations seems more like an abductive task, where there are often multiple hypothetical explanations for a set of sensor data. Just a suggestion in case inference over observation data fits into the scope of this project.  For reference please see: http://knoesis.wright.edu/library/resource.php?id=00595

On Tue, May 26, 2009 at 9:45 AM, Oscar Corcho <ocorcho@fi.upm.es<mailto:ocorcho@fi.upm.es>> wrote:
I tend to agree with Amit. Anyway, probably Víctor was not referring
necessarily to DL reasoning, but also to some RDF inferencing.

Oscar

-----Mensaje original-----
De: public-xg-ssn-request@w3.org<mailto:public-xg-ssn-request@w3.org> [mailto:public-xg-ssn-request@w3.org<mailto:public-xg-ssn-request@w3.org>] En
nombre de Amit P. Sheth
Enviado el: martes, 26 de mayo de 2009 14:23
Para: Victor Manuel Pelaez Martinez
CC: public-xg-ssn@w3.org<mailto:public-xg-ssn@w3.org>
Asunto: Re: another relevant bit of ontology

I would argue that DL inferencing is not the only or even primary way to
exploit semantics. We have had lits of debate in the community
and we generally agree that SW's benefits not only machine processing
but also, human understandability and use a of data. RDF processing is
perfectly
adequate ways to exploit semantics by exploiting relationships
as first class objects.
Semantic annotations can improve search, for example,
or integration, without using inferencing.
Some ideas are in this paper:
http://knoesis.wright.edu/library/resource.php?id=00050

Amit
http://knoesis.org/amit

Victor Manuel Pelaez Martinez wrote:
>
> Dear All,
>
> This is a quite interesting use case, but it's focused more on the
> sensor data than in the sensor network itself, so in our opinion it
> fits only partially in the scope of the group (as the group discussed
> in one of the teleconferences).
>
> We think that the most interesting point is the use of ontology
> reasoning (in this case using rules) because it goes further than the
> use cases based on syntactic queries.
>
> In our opinion we shouldn't be able to solve use cases using only
> syntactic queries, because that would mean that those use cases could
> be solved using other non-semantic technologies.
>
> We think that the use cases should show the advantages of using
> semantic technologies, so some kind of reasoning or inference should
> be necessary in order to solve them (as it is done in the paper sent
> by John). Perhaps this could be a good point to discuss within the use
> case work package.
>
> Regards,
>
> Víctor M. Peláez
>
> Fundación CTIC
>
> Departamento de I+D+i
>
> victor.pelaez@fundacionctic.org<mailto:victor.pelaez@fundacionctic.org> <mailto:victor.pelaez@fundacionctic.org<mailto:victor.pelaez@fundacionctic.org>>
>
> Tfno: +34 984 29 12 12
>
> Fax: +34 984 39 06 12
>
> Parque Científico Tecnológico Gijón - Asturias - Spain
>
> www.fundacionctic.org<http://www.fundacionctic.org> <http://www.fundacionctic.org>
>
> ------------------------------------------------------------------------
>
> *De:* public-xg-ssn-request@w3.org<mailto:public-xg-ssn-request@w3.org>
> [mailto:public-xg-ssn-request@w3.org<mailto:public-xg-ssn-request@w3.org>] *En nombre de *John Graybeal
> *Enviado el:* jueves, 21 de mayo de 2009 0:39
> *Para:* public-xg-ssn@w3.org<mailto:public-xg-ssn@w3.org>
> *CC:* Bob Morris; Matt Calder; Francesco Peri
> *Asunto:* another relevant bit of ontology
>
> Folks,
>
> Bob Morris and his team at UMass (copied) have a paper in review [1]
> that seems highly relevant as a demonstration use case. It describes
> using ontologies to relate types
>
> of ecosystem events to properties of the ecosystem and the sensors
> observing it.
>
> I added this reference to the Use Cases page, and also to the
> References page, but the sensor ontology itself may be of direct
> interest (I don't have a link to that yet).
>
> John
>
> [1] http://efg.cs.umb.edu/pubs/SensorDataReasoning.pdf
>
> --------------
>
> John Graybeal <mailto:graybeal@mbari.org<mailto:graybeal@mbari.org>> -- 831-775-1956
> Monterey Bay Aquarium Research Institute
> Marine Metadata Interoperability Project: http://marinemetadata.org
>



Se certificó que el correo entrante no contiene virus.
Comprobada por AVG - www.avg.es<http://www.avg.es>
Versión: 8.5.339 / Base de datos de virus: 270.12.39/2134 - Fecha de la
versión: 05/25/09 18:14:00





--
Cory Andrew Henson
Kno.e.sis Center, Wright State University
http://knoesis.wright.edu/researchers/cory/


John

--------------
John Graybeal   <mailto:graybeal@mbari.org>  -- 831-775-1956
Monterey Bay Aquarium Research Institute
Marine Metadata Interoperability Project: http://marinemetadata.org




--
Cory Andrew Henson
Kno.e.sis Center, Wright State University
http://knoesis.wright.edu/researchers/cory/


John

--------------
John Graybeal   <mailto:graybeal@mbari.org>  -- 831-775-1956
Monterey Bay Aquarium Research Institute
Marine Metadata Interoperability Project: http://marinemetadata.org
Received on Thursday, 4 June 2009 12:47:00 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 4 June 2009 12:47:01 GMT