[Top] [All Lists]

Re: [ontolog-forum] Quote for the day

To: ontolog-forum@xxxxxxxxxxxxxxxx
From: "John F. Sowa" <sowa@xxxxxxxxxxx>
Date: Tue, 04 Jan 2011 11:59:02 -0500
Message-id: <4D2351D6.9030308@xxxxxxxxxxx>
Anders, Pat, Mike, David, Ron, Toby, and Ed,    (01)

These discussions are beginning to show some convergence, but
more work is needed to fill in the details.    (02)

> What I was trying to convey was that a singular ontology approach
> is not likely to work. Instead it is most likely more beneficial
> to look for an ontology approach with interlinked micro ontologies
> and theories (MOT) that are 'pluggable' and adaptable (extend, restrict,...)    (03)

That is something I have been preaching for a long time:  the need
for modular ontologies and a systematic way of combining and relating
an open-ended number of interlinked micro ontologies (or microtheories
as they are called in some systems, such as Cyc).  Following are the
slides for a 3-hour tutorial in which I discussed such ideas:    (04)

    Integrating Semantic Systems    (05)

For a brief summary of the Cyc Project, see slides 13 to 15.    (06)

For more detail about the methods for relating an interlinked hierarchy
of micro ontologies, see slides 61 to 81.  (To jump to a particular
slide, just type the slide number into the window on the Adobe reader.)    (07)

> The above is an example of an eco-system view of ontologies.    (08)

I like that phrase.  Interoperability does not require a single
universal ontology, but an ecosystem of ontologies at different levels
of detail and with different ways of specializing that detail for
different purposes.  The ecosystem also requires systematic methods
for relating, detecting, negotiating, and resolving any differences
or conflicts that may arise.    (09)

>> I believe that a foundation ontology that contains an inventory
>> of all the known primitive elements would serve that function.    (010)

> This is an interesting notion to use higher 'level' elements
> to describe the similarities and differences in lower 'level' model.
> Next is to look at, how many 'levels' do you have? Primitives and Domain and 
>...?    (011)

I would replace Pat's term 'foundation' with 'ecosystem', and I would
support an open-ended (potentially infinite) number of "primitives",
any of which might be refined or redefined in terms of any others.    (012)

>> Of course, as new information is gathered the basic vocabulary itself
>> might also expand.  How quickly it expands, and how this might affect
>> existing applications is an empirical question that will be interesting
>> to explore.    (013)

I agree with this point.  But as soon as you include the option of
expanding a list of primitives, it's misleading to call them primitives.
My attitude toward primitives is one of tolerance:  I have no objection
to anybody's proposed primitives, as long as the list is open-ended,
and anybody can add as many primitives as they like at any time.    (014)

But I would also insist on definitional mechanisms (along the lines
of the lattices discussed in slides 61 to 81 of iss.pdf) that can
relate any terms (primitive or not) to one another.    (015)

In other words, the question about primitives is an empirical issue
that can only be resolved by the evidence about how and whether
any proposed set can be defined in terms of some smaller set.
In short, primitives are the *result* of practice and testing,
not an _a priori_ starting point.    (016)

> I'm interested in looking at large scale systems or large scale
> Eco Systems of interlinked models/MOT's.
> In this case the 'levels' usually  form longer chains of
> adaptation/derivations/transformations/interlinked MOT's.
> Most users pick up some Industry's (or CommunityOfPractice or
> IT vendor or...) end-of-chain and then continue with bilateral
> agreements and internal adaptations.    (017)

I agree with this approach.  I would recommend an open-ended
ecosystem in which anybody's ontology (vendor, user, or researcher)
can find a place.  Then new ontologies could be formed from the
existing ones by any kinds of agreements, adaptations, extensions,
contractions, or modifications.    (018)

The basic operators in the lattice of theories can support any chain
of revisions and combinations that relate or convert one theory to
any other.  See slide 79 of iss.pdf.  For more detail about the
operators in that slide, see pp. 16 to 25 of    (019)

    http://www.jfsowa.com/pubs/rolelog.pdf    (020)

The first 15 pages of this article present the philosophical
background, but it's possible to jump to page 16 for the operators
on the lattice without reading any of the philosophy.    (021)

> The financial industry does not lack precision. Not only simple
> monetary instruments (whose meaning is as primitive as it gets)
> but more complex financial instruments, whose meaning is grounded
> in contract and law.
> ... ISO 20022.    (022)

Following are two .ppt presentations that describe ISO 20022,
Universal financial industry message scheme:    (023)

    http://www.iso20022.org/    (024)

> The ambiguity sets in when various technical folks get hold
> of what are originally precise terms, and shoe horn them into
> various different databases with different data structures, put
> unexpected new (but precise) terms into existing data fields that
> also house data with different meanings, then devise message
> schemes that get these from place to place, all without formal
> reference to the original, existing, precise business meanings.    (025)

That is why we need an ecosystem with systematic operators for
relating interlinked -- call them what you will -- theories,
models, ontologies, specifications, or whatever.    (026)

> The fundamental issue here is there's no formal mechanism that attaches
> the business language & definitions to what gets written in code.    (027)

That is a very serious gap that must be filled (or bridged).  That is
the reason why I have been recommending controlled natural languages
supplemented with diagrams that are readable by the business types,
the systems analysts, and the programmers.  And those CNLs and diagrams
must be systematically mappable to *every* technology in the ecosystem
-- that includes the ontologies, the modeling tools, the negotiating
and contracting language, and all the implementation languages.    (028)

See slides 40 to 60 of iss.pdf    (029)

And by the way, I do *not* claim that writing controlled NLs is easy.
But CNLs (supplemented with good diagrams) are much easier to read
by *everybody* who is involved with the system -- from the business
end to the software implementations.    (030)

Without a common language for the business people and the programmers,
mistakes and misunderstandings are inevitable.  Even with a common
language, errors are going to be made -- but it becomes possible to
catch them and to explain in a readable way how & why they occurred.    (031)

> I think that there are opportunities now for some companies that have
> the vision to build useful tools that accelerate the process of
> compliance to the new government reporting requirements that are coming
> out in response to terrorism, the financial crisis and by an increasing
> desire in the part of politicians on all sides in all jurisdictions
> to talk about reducing wasteful regulation while wanting to control
> everything and the expectation that the various government agencies
> will monitor everything and instantly see patterns and individual
> transactions that pose a threat to society.    (032)

I strongly agree.    (033)

>> Not every business model *wants* precision of description and communication.    (034)

> That is a different matter.  The objective in a trade agreement is to be
> be as precise as necessary to get agreement and reveal as little as
> possible of the inner workings of each agent.  So it is not a matter of
> not wanting precision; it is a matter of choosing an appropriate precision.    (035)

I agree.    (036)

> A model of full and perfect knowledge on each side would slow
> business and reduce opportunity.    (037)

That gets into policy issues about how to use the tools.  Good tools
can do a lot to support good policies, but they won't replace people
(who will make good or bad decisions in their own inimitable ways).    (038)

> It is the culture of business administration people to think in terms
> of the paper form, rather than its information content.    (039)

We're not going to change their preferences anytime soon.  That's one
more reason why it's essential to focus on *semantics* -- and support
any and every syntax that any human or computer system uses.    (040)

>> Two trading partners adapt and agrees on their own adaptations, based
>> on their industry's Product' MOT    (041)

> This is the approach of the ISO/OASIS Product LifeCycle Services (PLCS)
> standards gang. They allow for successive levels of standardization,
> each of which defines common practices for a smaller industry group and
> allows for trading partner specializations and adaptations. Their
> models, however, are currently written in a combination of EXPRESS (with
> an OWL derivative) and XML Schema, so that they actually get implemented
> in commercial software.    (042)

Another argument for focusing on semantics and providing translations
to and from every notation on the planet.    (043)

>> The above is an example of an eco-system view of ontologies.    (044)

> After the abuse of this term in OMG and elsewhere, I don't know what
> an 'ontology eco-system' might be.  So if Anders says this is one,
> who am I to argue?    (045)

I admit that I never understood what that term meant in OMG,
but I like it better than Pat's term 'foundation'.  Since Anders'
view is close to my hierarchy of theories, I'm delighted to have
a catchy new term for what I've been talking about for years.    (046)

John    (047)

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (048)

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