comments (read: my opinions; your mileage might vary) below.
Jack (01)
paola.dimaio@xxxxxxxxx wrote:
> Hi John
>
<snip>
> Could the process of decoupling K from KR be called KAb?
> (02)
is "decoupling" being used in the same sense as "abstraction"?
see below... (03)
>> My only criticism is to note the statement in one of the
>> papers you cited: "knowledge representation and abstraction
>> are not independent."
>
> I am sure that sometime during discourse some mixup occurs, so not
> sure who said that
> in what context, and I think it has been well argued elsewhere that
> abstraction and representation are two different phases of the
> knowledge cycle.
><snip> (04)
So let's have a go at a definition: (Not even Abbot has come up with
> a neat one yet, that I can find)
>
> Knowledge Abstraction is the process of
>
> a) representing K independently from formalism
> b) capturing K in a form that cane be easily represented by diferent
>formalisms
> c) representing K with the minimal and least formalised representation
>
> or?
> (have your say) (05)
Huh?
I went to http://en.wikipedia.org/wiki/Abstraction, a page that almost
looks like it has John Sowa's footprints all over it (conceptual graph
of cat on mat). There, the statement is made (among many):
> In philosophical terminology, abstraction is the thought process wherein
>ideas are distanced from objects. (06)
In statement a) above, it appears to me that the idea (K) is being
distanced from the representational schema (my interpretation of
"formalism") as compared to objects being abstracted. Would that be a
new way to use the term "abstraction"? (07)
It seems to me that there are at least three different subjects in play
here:
<concepts/ideas/entities/...>,
<knowledge representation implementations>,
<abstractions as a the results of processes> (08)
It strikes me that I can abstract Paola DiMaio to instanceOf Person,
throwing out lots of information *about* Paola. That's just a thought
process. I can then represent that abstraction in a tiny taxonomy, the
likes of which can be implemented in xml, rdfs, owl, you name it... (09)
I think I get that the intent of OpenOntology was to capture knowledge
(whatever that is) in some implementation-neutral way. I earlier
suggested topic maps as being, um, precise enough to capture what is
observed/thought/known/believed and still be able to be exported to any
formalism/KR scheme desired. That suggestion got no traction here. (010)
>
> My bottom line is: in order to enable knowledge to be accessible and
> readily usable by different systems and architectures, it should exist
> as independent as possible from the constraints of a formalisation
> (could be any formal language, or OWL for example),
> (probably more to say there)
>
> Then is the question , how can K be expressed with precision in its
> abstract form?
> (011)
Just how precise would you have the class Person be? (given Person is an
instance of an abstraction of, say, Paola DiMaio). Would you choose a
different, um, abstraction? To serve which purpose? Indeed, how do you
define "precision"? (012)
> I suspect Natural Language and XML structure is the first thing that
> comes to mind,
>
> That's my claim for semantic technologies potential to support
> knowledge based reasoning in open, dynamic environments
>
> Or maybe mindmaps? (mind map concept is highly abstract in itself) (013)
Why does "mindmaps" come to mind here? They are simply concept maps
(nodes with labeled arcs) with a particular twist: there is always a
central concept in view. When you represent knowledge with simple
structures such as that, you give up being able to treat the
relationship instances (the labeled arcs) as subjects (nodes) in their
own right. It strikes me that conceptual graphs make a more, um, precise
way to capture your thoughts. Yes indeed! Mind maps abstract away all
the fun (read: important/precise) stuff! (014)
Since I wear many hats, one of which says "topic maps", I'll point out
that a conceptual graph, in my judgment, can serve as a fine topic map
so long as you pay attention to representations of subject identity in
each node. (015)
> But also UML?
> Are ER diagrams and inference networks the most abstract form of K
> that we are capable of to date?
>
>
> Not sure if I see the knowledge abstraction to be the role of a single
> person, I see it more as a process of distillation, the result of a
> method that needs to be developed maybe, the knowledge development
> lifecycle I think, of which abstraction could be the last refinement
>
> I am sure there is a lot to argue, and a lot to unpack, and a lot of
> possibilities there
>
> Tomorrow I ll put it in the wiki for the Summit perhaps, and wait to
> hear what you have in mind.
>
>
> Paola Di Maio
>
>
> On 2/7/07, John F. Sowa <sowa@xxxxxxxxxxx> wrote:
>> Paola,
>>
>> I very strongly agree that abstraction is extremely important
>> and that it should be recognized as a valuable technique
>> that is applicable to subject matter of any kind.
>>
>
>> People have been using abstraction techniques together
>> with knowledge representation since the time of Aristotle.
>> To avoid multiplying different fields, I suggest that one
>> term be used as the overall umbrella word that covers all
>> the others.
>>
>> I really don't care whether the overall term is called
>> "knowledge representation", "conceptual analysis", or
>> whatever. But it's important to have an umbrella term
>> for all the cases. Then you can have parallel terminology:
>>
>> knowledge representation -- the task of analyzing the
>> concepts in any body of knowledge, defining appropriate
>> abstractions, and mapping the results to a logic-based
>> notation.
>>
>> knowledge engineer -- one who does knowledge representation.
>>
>> If you have too many terms at the top level, you get a
>> combinatorial explosion:
>>
>> knowledge engineer -- one who does KR
>>
>> abstraction engineer -- one who abstracts
>>
>> ontological engineer -- one who develops ontologies
>>
>> conceptual engineer -- one who analyzes concepts
>>
>> And all the possible combinations:
>>
>> knowledge & abstraction engineer
>>
>> ontology & abstraction & conceptual engineer
>>
>> knowledge & ontology & abstraction engineer
>>
>> etc.
>>
>> If the same person is expected to do all these things, we need
>> a single term for the subject and the person who works on it.
>>
>> John Sowa
>> (016)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx (017)
|