[Top] [All Lists]

Re: [ontolog-forum] Reality and semantics. [Was: Thing and Class]

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "Len Yabloko" <lenya@xxxxxxxxxxxxx>
Date: Thu, 18 Sep 2008 03:55:55 +0000
Message-id: <W580927735186171221710155@webmail14>
>On Sep 17, 2008, at 12:30 PM, Len Yabloko wrote:
>> Pat,
>> As you concluded this very intellectually challenging debate with  
>> John, and two of you finally agreed to disagree, - let me see what  
>> it means for software developer like myself.
>Sure.  But might I ask, what kind of software you develop? How does it  
>- the software - relate to ontologies?    (01)

I develop software for Electronic Health Records and aspire to expand into more 
general medical knowledge management applications. But I think all 
data-intensive software relates to ontologies in more fundamental way than what 
you call 'mathematical language' in broad sense.     (02)

I agree with John and others who pointed out that information compression is a 
true subject of this discussion - not language. In that view - language is one 
of the facets supporting information compression, and in that capacity it does 
not have an independent value outside narrow mathematical sense.     (03)

This is why natural and programming languages do not relate directly, or even 
at the level of model theory - their scope of information compression is 
completely different. And this is also why micro-theories or as someone put it 
- "network of theories" is where software and ontologies meet. In software 
parlance ontology is called "interface" - the place where information is 
compressed and decompressed. And, as I am sure you know,- direct interfacing 
leads to "n-square" problem (see my webside http://www.ontospace.net ).     (04)

Mediation, on the other hand, requires interpreation of data as opposed to 
commands. This is where RDF, OWL and other model-theoretic semantics enter the 
world of software. Unfortunately, as John pointed out these semantics are not 
compatible with database semantics used to process application data.    (05)

>>> On Sep 16, 2008, at 7:30 PM, John F. Sowa wrote:
>>>> Pat and Chris,
>>>> JFS>>> A Tarski-style model is a set of entities and a set of
>>>>>>> relations among those entities.  Those entities and relations are
>>>>>>> *approximate* representations of aspects of the world according  
>>>>>>> to
>>>>>>> some ontology.
>>>> CM>> Sure.  Every model leaves out information that is found in the
>>>>>> piece of the world that the model represents.
>>>> PH> Represents? Ouch. If there is a 'represents' relationship  
>>>> between
>>>>> models and reality, then all our axiomatic ontologies must be given
>>>>> a two-stage semantics, in which model theory describes the first
>>>>> stage of interpretation, yielding a new kind of 'representation'
>>>>> which then needs another, presumably different, semantic theory to
>>>>> relate it to actuality.
>>>> Welcome to reality.  As George Box said and I quoted in my previous
>>>> note, "All models are wrong, but some are useful."
>>> Its not reality. In fact, we all manage extremely well with a single-
>>> layered semantics called model theory.
>> Most of software developers would probably say: "we all manage  
>> extremely well without it", and many of us might add "without any  
>> semantics at all".
>I very much doubt that. I don't think it is possible write any  
>nontrivial software without having some command of semantics, often a  
>fairly precise one. Of course it may not be expressed mathematically  
>or in terms of model theory by those who use it, but nevertheless it  
>has to be there. And when one looks at the amount of effort devoted to  
>getting software well-structured in one sense or another, the  
>importance of a precise semantics seems to loom increasingly large, at  
>least it seems so from what I have read. Certainly many of the common  
>mistakes made by beginning developers of RDF and OWL systems stem  
>directly from a failure to grasp the semantics of these formalisms.
>> Few of us may note that we only need operational semantics. But none  
>> of us will say what you said. So I don't know who do you refer to as  
>> "we all", and what do you call to "manage extremely well".
>I was meaning to refer to people who seek to give semantic theories  
>for assertional formal languages used in ontologies, actually, rather  
>than software developers.
>>> This two-level idea is a
>>> chimera, and an intellectual dead end.  You have argued for the idea,
>>> but do you have even an outline or a sketch of what the second,  
>>> model-
>>> to-reality, semantic theory looks like? In order to give it, you need
>>> to somehow describe reality mathematically. In full generality. OK,
>>> I'm all ears.
>> I would say that most people manage extremely well without  
>> "describing reality mathematically".
>In a broad sense of 'mathematical', I would say that virtually all  
>software describes reality mathematically.     (06)

This is where I think you missed the point that John was making about 
real-world problems not addressed by 'mathematical semantics'. You see - 
describing reality is not an objective in itself, certainly not for business 
application software, interacting or manipulating it - is. To do it well 
requires efficient compression of information. In that sense theorem proving is 
irrelevant (this kind of semantics are only required for proof-carrying code).     (07)

>But in this context, we  
>were referring to (hopefully reasonably precise) semantic theories,  
>which are pretty much required to be mathematical, in the sense that  
>they are couched in sufficiently formal language as to enable one to  
>prove theorems in them.
>> In fact most believe that it is impossible to do so.
>>>> PH> Not only has this project never been undertaken to completion,
>>>>> I don't think its needed.
>>>> That is why 20th century analytic philosophy has trivialized the
>>>> subject to the point where the total number of bookshelves devoted
>>>> to philosophy in a large Barnes & Noble store is equal to the
>>>> number devoted to Sudoku puzzles.
>>> This comment is completely off the wall.
>> I think it is literally "off the wall" of any profitable book store.  
>> Or else it would quickly go out of business. Because the semantics  
>> you are finding so useful are actually totally useless for real  
>> world that they..., should I say represent ?
>>> This is such rubbish that I won't even try to sketch a response.  
>>> Which
>>> *semantic* problems are being ignored?
>> I'd like to ask what *semantic* problems are NOT ignored by software  
>> community. Certainly Semantic Web community ignores most of them  
>> beginning with most basic identity problems.
>Well, do tell us what these semantic problems are. Give an  
>illustrative example.    (08)

As John already pointed out the most acute problem is identity of instances 
compressed and decompressed during communication. You alluded that some of us 
can't tell "hash table from shopping list". Then may be you can help us to 
figure out how shopping list can be passed around between different systems and 
still be useful if hash code is calculated locally by each system? What 
semantics do you propose us to use in order to establish equivalency between 
substitutable items from different manufacturers?    (09)

>    (010)

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    (011)

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