[Top] [All Lists]

Re: [ontolog-forum] Ontology and methodology

To: Florian Probst <f.probst@xxxxxxxxxxxxxxx>, "Cassidy, Patrick J." <pcassidy@xxxxxxxxx>
Cc: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: Pat Hayes <phayes@xxxxxxx>
Date: Wed, 21 Mar 2007 10:56:15 -0500
Message-id: <p06230903c226f951b40c@[]>
>>>  What is the purpose of a foundational ontology if
>>>not to serve as common denominator? Or in your words, as a set of
>>>concepts out of which you can define all others?
>>Well, yes. Do you think you can use DOLCE to *define* all other 
>>concepts? I would be amazed if you could.
>no, of course not. The foundational ontology provides an entry point 
>for introducing new (domain level) concepts with which one is able 
>to make a certain conceptualization explicit. If the foundational 
>ontology would allow to define all concepts exhaustively, then the 
>number of concepts and the number of conceptualizations would be 
>fixed, which is obviously not a good idea.    (01)

Take this up with Pat Cassidy. I believe he really does mean this by 
the phrase "foundation ontology". He does not mean what has been 
traditionally called an "upper ontology". In his vision (correct me 
if I get this wrong, Pat), the number of  defined concepts would not 
be fixed, because one could combine foundational 'primitive' concepts 
in various ways.    (02)

>In contrast, if a domain ontology engineer commits to a foundational 
>ontology, some other domain ontology engineer can grasp the general 
>lines of conceptualization.    (03)

You are referring to what I was calling a 'framework'. It amounts to 
an agreed metaphysics, in effect: agreeing to adopt a certain 
perspective on certain basic issues, often having to do with 
relationships and time and necessity, and to use a certain basic 
style of description when talking of these things. DOLCE is in this 
category. In your terminology, what PatC is proposing is (I believe) 
more like a kind of universal domain ontology construction set. I 
would place CYC in this category, and so far its about the only one 
of its kind. And this is what I was suggesting can be 'grown' by 
using the semantic web.    (04)

>  The foundational ontology provides the "foundation" for making a 
>specialized conceptualization explicit. Of course new concepts needs 
>to be added in order to arrive at a domain ontology. If concepts are 
>added that carry the same name, yet are attached to different 
>foundational ontology concepts differences in the conceptualization 
>can easily be tracked.    (05)

That is often claimed to be the point of such frameworks and upper 
ontologies, but it is rarely argued for, usually just assumed, and I 
remain to be convinced. I think that experience bears me out. No 
matter how elaborate the explanations of the framework, people 
quickly start using it in incompatible ways. (If, like DOLCE, it 
requires one to distinguish between continuants and noncontinuants, 
then there will be debates over whether or not something is a 
continuant: in many cases it is very hard to decide. Such debates 
have already happened.) Seen from the other direction, the actual 
content in large ontologies often seems to be rather loosely 
connected to the top-level distinctions, and largely independent of 
them. That has been the CYC experience, certainly: Doug Lenat often 
says that while one needs an upper ontology for computational reasons 
(to be the top levels of the subsumption/inheritance hierarchy), the 
particular choice is largely unimportant, and the great bulk of the 
content of a domain ontology can be quickly morphed so as to fit 
underneath almost any 'upper' classification. It is smart 
engineering, in fact, to do the details of upper parts *last*, as a 
byproduct of the process of writing domain ontologies and negotiating 
how to put them together. Similarly, the lowest levels amount often 
to little more than data. All the actual hard ontological work is 
done in the middle sections, what would be the upper reaches of a 
particular domain ontology. BTW, the other great lesson from CYC is 
that things get a lot, lot easier if one is willing to allow multiple 
'points of view' (aka 'contexts' aka 'microtheories') within a single 
ontology, rather than trying to get all the distinctions right before 
starting; this often removes the need to get a lot of uniform global 
consistency between different subareas.    (06)

>>>I think, Patrick C made
>>>an important point by asking how to relate the many different formalized
>>>conceptualizations (domain ontologies) if not via a small set of concepts.
>>By writing axioms and/or translation rules which do the relating, 
>>is how. It takes work, and a degree of willingness to immerse 
>>oneself in alien ways of thought, but it can be done.
>I suppose that developing a domain ontology serves
>1) to make a certain conceptualization (world view) explicit
>2) to provide meaning to symbols that are used to denote those 
>entities that are instances of the concepts made explicit in the 
>I am only sure about the meanings of symbols in my own ontology. How 
>should I know what the symbols in your ontology denote, if I have no 
>(!) agreed on entry point for understanding your conceptualization?    (07)

You read my ontology :-) Most of them are fairly self-explanatory if 
displayed properly, and if you know how to not misinterpret the 
underlying formalism or logic they are written in. (BTW, for OWL, try 
viewing it COE.) If you really cannot understand mine at all, you 
probably shouldn't be using the concepts in it. That is OK, nothing 
will break if you re-invent my wheel. Maybe someone else will provide 
a fishplate ontology to translate between us later.    (08)

The more complex the formalism, the harder they are to read, I will 
concede. CYC is a bear. But on the other hand, such very complex 
ontologies will usually have specialized viewers, etc., since their 
own authors will have needed them.    (09)

>How do you make sure that the axioms and/or translation rules you 
>are proposing do justice to my ontology?    (010)

Im not sure what you mean by 'do justice to'. I may use concepts from 
your ontology in ways that you never intended, of course. But this is 
part of the social process of growing a mutual understanding. This is 
already happening in the SWeb: my favorite example is the Dublin Core 
notion of 'author' which the DC writers understood in a librarian's 
sense, but which is now, thanks to its use in the FOAF project, far 
more widely used to relate composers of Web pages or email messages 
to their electronic products. Of course this is *consistent* with the 
original intention, but it goes beyond it. Is this 'doing justice'?    (011)

>It seems not appropriate to rely on natural language semantics for 
>translating between ontologies, since it is the purpose of the 
>ontology to disambiguate the natural language semantics.    (012)

That is one purpose, but there are many others. I wouldn't say it was 
the main one.    (013)

>Can you translate between two ontologies where all "words" are 
>(consistently) encrypted?    (014)

I'm not sure what your point is here.    (015)

Pat    (016)

>You can match symbols, but I guess you can not assure that the 
>symbols keep their meanings. But isn't the possibility to negotiate 
>meaning the central benefit of building ontologies?
>>>For geographic space, any country has its own geographic coordinate
>>>system and specialized map projections. The world that these maps
>>>represent is in all maps the same, yet it might be differently depicted
>>>(represented). In cartography, coordinate transformation is a long
>>>solved problem, because it was possible to define a "geodetic datum".
>>>If  two cartographers commit to such a geodetic datum, they can
>>>translate the positions represented in one map(projection) into another.
>>>Why not achieving such a "semantic datum" for ontologies?
>>Briefly, because there is no reason to suppose that it is possible, 
>>and lots of reasons to suppose that it is impossible. As I said in 
>>an earlier email to (I think) Paula, conceptual "space" isn't like 
>>physical space. We don't even know if it has a meaningful topology, 
>>let alone a metric geometry (which was required for the 
>GI-Days 2007 "Young Researchers Forum": http://www.gi-days.de    (017)

IHMC            (850)434 8903 or (650)494 3973   home
40 South Alcaniz St.    (850)202 4416   office
Pensacola                       (850)202 4440   fax
FL 32502                        (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes    (018)

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    (019)

<Prev in Thread] Current Thread [Next in Thread>