[Top] [All Lists]

Re: [ontolog-forum] Foundation ontology, CYC, and Mapping

To: "[ontolog-forum]" <ontolog-forum@xxxxxxxxxxxxxxxx>
From: "doug foxvog" <doug@xxxxxxxxxx>
Date: Sun, 7 Mar 2010 12:27:27 -0000 (GMT)
Message-id: <3760.>
John Sowa wrote:    (01)

> Doug,    (02)

John, we agree.    (03)

> A formal system, such as the lattice of theories, is designed to
> support automatic mapping between theories in the lattice.  But
> I strongly doubt that arbitrary ontologies could be automatically
> with any degree of precision.    (04)

We agree.  An automatic mapper could be beneficial in creating an initial
alignment of two ontologies.  For an accurate mapping the result would
have to be vetted, with sub/super/closely overlapping meanings being
determined & the creation of rules distinquishing the differences.    (05)

> DF> One has to consider the purpose of a given ontology.  For many
>  > uses the exact boundaries of concepts are not important.  A company
>  > selling products -- furniture, tools, electronic goods, etc. --
>  > will be selling a set of specific (often brand-name) products, but
>  > define them as members of broader categories (chair, electric drill,
>  > telephone) for which there are ontological terms defined in different
>  > ontologies whose exact boundary conditions vary.  The locally defined
>  > terms would be subclasses of any of the similar, but not precisely
>  > identical, more general classes of products.    (06)

> Yes. Those are the kinds of vague mappings that search engines do,
> and they can be very useful for information retrieval.  But they
> can't support precise reasoning, much less interoperable software
> design and development.    (07)

Merely defining subclass relationships can not do this.  However,
although asserting that an AcmeModel123_RockingChair is a subclass of
cyc:RockingChair, dbp:RockingChair, and gr:RockingChair would not allow
reasoning that a specific instance of fo:RockingChair_Type7 is an instance
of AcmeModel123_RockingChair; it would allow the converse reasoning.
If enough features of the ACME model could be defined to allow the subclass
conclusion, then the rules defined for concluding an instance of
fo:RockingChair_Type7 is an Acme Model 123, could be automatically
modified to allow the same conclusion for instance of any of the other
slightly varying defined classes of rocking chair with the specifed
features.    (08)

My point was that in many ontologies, the non-leaf nodes are fuzzy.  They
are merely there to support the leaf nodes, which have crisper meanings.
Defining the non-leaf classes as merely the disjunction of all the defined
leaf subclasses is not appropriate, as that is not the "intent" of the
ontology designer.    (09)

> Some people claimed that they have "aligned" their ontologies
> to WordNet.  That can be useful for the vague mappings used
> by search engines.    (010)

> But WordNet is not a formal ontology, and the WN terms have no
> formal definitions.  The fact that two ontologies, A and B,
> have "aligned" some of their categories with some synsets in
> WordNet does not imply that corresponding terms in A and B
> have the consistent definitions.    (011)

Agreed.  Some WN terms have clear, although informal meanings, while
most do not.  A number of WN synsets are problematic.  If both
ontologies have only mapped to clear synsets (a big if), there is a strong
likelihood that the corresponding terms in the two ontologies have
consistent definitions    (012)

> DF> One advantage of an FO would be that people who want an ontology
>  > would be likely to use appropriate parts of the existing FO than
>  > to create a new ontology from scratch.  They may create new
>  > subclasses (new subtypes of existing products), or more specialized
>  > relations, but use the FO for the majority of their ontology.    (013)

> I agree that such developments can be very useful.  But a library
> of modules would be more useful for that purpose than a large,
> fixed ontology.    (014)

I was considering the FO as modular, above.  I was pushing that earlier,
and had understood that that was the consensus of the discussion.    (015)

-- doug    (016)

> John    (017)

doug foxvog    doug@xxxxxxxxxx
=============================================================    (018)

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    (019)

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