[Top] [All Lists]

Re: [ontology-summit] Criteria for evaluating ontologies at different le

To: ontology-summit@xxxxxxxxxxxxxxxx
From: John F Sowa <sowa@xxxxxxxxxxx>
Date: Fri, 28 Dec 2012 12:07:11 -0500
Message-id: <50DDD1BF.8070302@xxxxxxxxxxx>
Dear Matthew,    (01)

The main point I want to emphasize is the need for examples to clarify
the discussion.  It's impossible to explain any abstract points without
concrete examples -- and the best examples come from implementations.    (02)

> I have developed an integrated "upper ontology" for all of Shell's
> downstream business, which covered what you describe and more.    (03)

That would be an excellent example.    (04)

> My experience at Shell confirms your [MFU's] experience of between
> 1000-2000 concepts for an upper ontology with the scope you describe.    (05)

Can you post that ontology on the wiki or other web site?  If it has
some proprietary features, could you remove them before posting it?    (06)

>> The sales model, the engineering model, and the manufacturing model
>> will require very different upper ontologies...
>> They would also require very different detailed layers.    (07)

> More interesting (and important) is what they share.    (08)

Both are important, but we need examples.  Can you provide them?    (09)

> When Shell first tried to develop common data about the performance
> of its UK downstream business it found more than 100 different coding
> systems for its products.    (010)

That is typical of most enterprises.  The problem is even more serious
when independently developed software at different companies must
interoperate.  That is why the Amazon.com example is important.    (011)

>> I agree that systems analysts have a lot to learn about ontology.  But
>> most academic philosophers and logicians are totally clueless about how
>> to design a working system.  Both kinds of skills are important, and
>> very few people have much, if any experience with both.    (012)

> And that is one of the biggest problems.    (013)

>> We desperately need good tools that bridge the gap between the two.    (014)

> No, we need education. A fool with a tool is still a fool.    (015)

The domain experts or SMEs are not fools.  They have years of practical
experience, and many of them have an MS, PhD, or MD.  It would be more
foolish to waste their time in a course on philosophical jargon.  It's
far better to design tools based on the notations and terminology that
they use to communicate among themselves.    (016)

>> The ontology tools are woefully inadequate for system design and
>> development, and the tools for systems analysis don't provide any
>> guidance about ontology design and use.    (017)

> Nor should they. In my view it is important for tools to be ontological
> commitment neutral.    (018)

I agree that the tools themselves and the logic that supports them
should be as neutral as possible.  But you need to load the tools with
a particular ontology, the words that express it, and the mappings to
and from the notations that the domain experts use.  As soon as you do
that, the tools are ideally adapted to what the domain experts know.    (019)

JFS, quoting and modifying an article on Big Data
>> What if domain experts could directly encode their ideas and
>> representations of their domains into the system, bypassing the data
>> scientists [ontologists] as middleman and translator?    (020)

> As I recall, that was the claim made for SQL..., and for COBOL as the
> "end of programming".    (021)

COBOL was designed over 50 years ago.  SQL was designed nearly 40 years
ago.  I won't claim that they're perfect.  But they certainly got far
more practical use than any notation for ontology or the Semantic Web.    (022)

> You are extremely unlikely to produce a good ontological model without
> a good working knowledge of ontological principles and problems.    (023)

The people who design the tools certainly need that knowledge.  But it
is possible to extract a proto-ontology directly from documents written
by the domain experts in ordinary NLs.  Then the domain experts can
update and correct the proto-ontology using only terminology and
notations that they prefer.    (024)

We've developed such tools at VivoMind, and the users love them.
But we're not the only ones who are working on these issues.  For
an overview see    (025)

    Future directions in semantic systems    (026)

John    (027)

Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/   
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/  
Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx
Community Files: http://ontolog.cim3.net/file/work/OntologySummit2013/
Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2013  
Community Portal: http://ontolog.cim3.net/wiki/     (028)
<Prev in Thread] Current Thread [Next in Thread>