RE: [URW3 public] OWL extensions [was Re: [URW3] ... three questions based on the last telecon]

> -----Original Message-----
> From: Peter Vojtáš [mailto:Peter.Vojtas@mff.cuni.cz]
> Sent: Wednesday, July 25, 2007 8:15 PM
> To: Giorgos Stoilos; 'Ken Laskey'
> Cc: 'Kathryn Blackmond Laskey'; public-xg-urw3@w3.org; 'Umberto Straccia'
> Subject: RE: [URW3 public] OWL extensions [was Re: [URW3] ... three
> questions based on the last telecon]
> 
> Thanks to Giorgios, we converge slowly.
>      I do not say all computations are running localy. E.g. for top-k
> query we have developped a model of web service where a B+ tree is
> indexing entries and the fuzzy set sent to it navigates through the tree
> (see our lecture at Fuzz_IEEE).
>      What I do say are two things:
> 1. let us specify more detailed what is in which form on the web and what
> is localy at user at the time of the query
> 2. Second (implied by this) if we have two sources and and one puts 178
> tall in degree 0.6 and the second source puts 180 tall in 0.5, then I have
> problem.

Right. But in my opinion the best solution will not be to put things on the
user side. Cause as I said in my preview mail this implies, 
i) intelligence of the end-user to create ontologies, fuzzifiers, map his
queries to web sources or services etc (will hardly ever happen) and 
ii) high computational efficiency of end-users system. 

Even if you have the above 2 the problem might again not be solved since the
realistic thing is that source1 and source2 have an ontology by which they
publish their content on the (semantic) web. So the one ontology could have
the axiom Tall = \exists hasHeight.>=178, while the second one Tall =
\exists hasHeight.>=185, so you still have a problem. In fact this will be
worse because you will get 178 as a result while you will not get 180 at all
(even as a lower rank).

In my opinion the best solution would be to use techniques like the ones
Umberto mentioned, like fusion or ontology alignment, to integrate the
results or source1 and source2, and leave user with the only difficult thing
to click a button :).  

>       I think even in use cases we have to be specific, in which form are
> data (annotated, structured, text), from different sources,...how to
> integrate data (not speaking about trust) what about user ...

As for this point I fully agree that Use Cases must be as specific as
possible.

-gstoil

>      Peter
> 
> 
> Please note my changed address Peter.Vojtas@mff.cuni.cz
> 
> 
> 
> ----- Original Message -----
> From: Giorgos Stoilos [mailto:gstoil@image.ece.ntua.gr]
> To: Peter.Vojtas@mff.cuni.cz, 'Ken Laskey' [mailto:klaskey@mitre.org]
> Cc: 'Kathryn Blackmond Laskey' [mailto:klaskey@gmu.edu], public-xg-
> urw3@w3.org, 'Umberto Straccia' [mailto:umberto.straccia@isti.cnr.it]
> Subject: RE: [URW3 public] OWL extensions [was Re: [URW3] ...  three
> questions    based on the last telecon]
> 
> 
> > Hi Peter,
> >
> > There are quite some drawbacks of this "localized computational" model.
> >
> > So, a user queries "get me all tall people". Now in order for the system
> to
> > return the exact height of each person and then a local processing
> method to
> > fuzzify the heights the following should hold.
> > 1) The (local) system should rewrite the query to "get me all people +
> their
> > heights"
> > 2) The (local) system should fuzzify the returned heights (according to
> user
> > preferences).
> > 3) The (local) system should apply a (probably) heavy reasoning
> algorithm to
> > reason with the concept tallness (in case Tall is a concept in some OWL
> > ontology).
> > 3b) Te last also might imply that the user has created (or imported from
> the
> > web) an ontology about people (heights, weights, ages, etc.)
> > All the above suggest a) much computational complexity on the user side
> and
> > b) that the user is experienced and capable enough to create
> fuzzification
> > functions and finally import or create OWL ontologies.
> >
> > After some thinking I do not see what is the problem on someone making
> for
> > me the decision that 178cm is Tall to degree 0.6. When a user (at least
> from
> > my point of view) queries "get me all tall people" he expects to get all
> > tall people in some descending order. So John who is 178cm will be
> returned
> > (say) 5th with degree 0.7, as he will also be returned 5th when using
> some
> > different local but homomorphic fuzzification function.
> >
> > In any case to my point of view the important thing is not to say to
> which
> > degree 178cm is tall, but to reason wrt fuzziness and return an order
> set of
> > results, reflecting users intuition.
> >
> > -gstoil
> >
> >
> > > -----Original Message-----
> > > From: public-xg-urw3-request@w3.org [mailto:public-xg-urw3-
> request@w3.org]
> > > On Behalf Of Peter Vojtas
> > > Sent: Wednesday, July 25, 2007 2:09 PM
> > > To: Ken Laskey
> > > Cc: Kathryn Blackmond Laskey; public-xg-urw3@w3.org; Umberto Straccia
> > > Subject: Re: [URW3 public] OWL extensions [was Re: [URW3] ... three
> > > questions based on the last telecon]
> > >
> > >
> > > Example is nice, but I would stress to use web examples, even if it is
> in
> > > an article on the web, somebody says ...XY...is tall, but he/she does
> not
> > > claim tall:0.7, we have to tend to more realistic web-examples (e.g.
> in
> > > our use cases). I agree it depends on the society, circumstances,....
> are
> > > we going to model also these? Peter
> > >
> > >
> > > Please note my changed address Peter.Vojtas@mff.cuni.cz
> > >
> > >
> > >
> > > ----- Original Message -----
> > > From: Ken Laskey [mailto:klaskey@mitre.org]
> > > To: Peter.Vojtas@mff.cuni.cz
> > > Cc: Kathryn Blackmond Laskey [mailto:klaskey@gmu.edu], public-xg-
> > > urw3@w3.org, Umberto Straccia [mailto:umberto.straccia@isti.cnr.it]
> > > Subject: Re: [URW3 public] OWL extensions [was Re: [URW3] ...  three
> > > questions   based on the last telecon]
> > >
> > >
> > > > Peter,
> > > >
> > > > We would always like to have "exact" information, but any
> measurement
> > > > has a degree of inexactness/uncertainty based purely on the
> preciseness
> > > > of the measuring instrument.  At a crime scene, a suspect may be
> > > > described as tall but the accuracy depends on what the observer
> > > > considers tall, the vantage point from which the observer saw the
> > > > suspect, and whether the suspect was wearing shoes with heels.
> Also,
> > > > was the suspect tall in the context of a society where the average
> > > > height is 162 cm or 175 cm?  Making use of the fact that the suspect
> > > > was "tall" requires many assumptions, most of which are implied in
> > > > conversation but never explicitly stated.
> > > >
> > > > What does this tell us about what needs to be represented in terms
> of
> > > > uncertainty?
> > > >
> > > > Ken
> > > >
> > > > On Jul 24, 2007, at 9:29 AM, Peter Vojtáš wrote:
> > > >
> > > > >
> > > > > I personaly would prefer to know the exact height of John and
> decide
> > > > > on my background and intention whether he is or not tall.
> > > > >      So I am afraid that I do not understand where such an
> information
> > > > > can appear
> > > > > Peter
> > > > >
> > > > >
> > > > > Please note my changed address Peter.Vojtas@mff.cuni.cz
> > > > >
> > > > >
> > > > >
> > > > > ----- Original Message -----
> > > > > From: Kathryn Blackmond Laskey [mailto:klaskey@gmu.edu]
> > > > > To: Umberto Straccia [mailto:umberto.straccia@isti.cnr.it],
> > > > > public-xg-urw3@w3.org
> > > > > Subject: Re: [URW3 public] OWL extensions [was Re: [URW3] ...
> three
> > > > > questions  based on the last telecon]
> > > > >
> > > > >
> > > > >>
> > > > >>
> > > > >>>> .... you can extend the language and the inference mechanism or
> > > > >>>> express and process the uncertainty within the standard
> language.
> > > > >>>>
> > > > >>>> tall(John) : 0.7
> > > > >>>>
> > > > >>>> vs
> > > > >>>>
> > > > >>>> tall(John, 0.7)
> > > > >>>>
> > > > >>>> (... in both cases, without saying what 0.7 represents)
> > > > >>
> > > > >> Independent of which way we go on tall(John) : 0.7 or
> tall(John,0.7),
> > > > >> it will not be enough just to annotate sentences with a number
> > > > >> expressing some degree of certainty or plausibility or membership
> or
> > > > >> whatever.  To do probabilistic reasoning, we need to be able to
> make
> > > > >> conditional independence statements, and to express conditional
> > > > >> probabilities. To do probability tractably depends on
> representations
> > > > >> composed out of local modules, and these local modules are
> > > > >> parameterized by conditional probabilities, not absolute
> > > > >> probabilities.
> > > > >>
> > > > >> K
> > > > >>
> > > > >>
> > > > >
> > > > >
> > > > --------------------------------------------------------------------
> ----
> > > > -----
> > > > Ken Laskey
> > > > MITRE Corporation, M/S H305      phone: 703-983-7934
> > > > 7151 Colshire Drive                         fax:       703-983-1379
> > > > McLean VA 22102-7508
> > > >
> > > >
> >
> >
> >

Received on Thursday, 26 July 2007 09:45:30 UTC