Re: Status of SUMO?
Philippe,
What is common or familiar depends upon your background. Folks who are
used to Davidsonian case role formalizations in linguistics and AI folks
who are used to instance and subclass statements will find SUMO's argument
ordering consistent with their experience. I would agree that consistency
is a good thing in general, and that a typical paraphrase into English
would result in a different word order than predicate argument order, but
that doesn't seem to me to outweigh either the inertia of the current
state, or the conventions of the communities I've mentioned.
For example
(agent ?WALKING John) John is the agent of walking
(instance ?WALKING Ambulation) ?WALKING is an instance of ambulation.
are fine by the principles I've mentioned. If you find other cases where
argument order is not consistent though, I'd welcome specifics. I recall
discussions with Ian on one or two relations where I would have preferred a
different argument order, but this sort of thing is really not much more
significant than the names of terms. We need to help the human user as
much as possible, but the primary focus of a formal ontology has to be on
getting the axioms right (in this case for first order inference).
Just to clarify, 'instance' and 'subclass' are simple first-order
relations, since they take terms, rather than sentences, as arguments.
Adam
At 05:36 PM 6/9/2004 +1000, Philippe Martin wrote:
>Adam,
>
> > SUMO has been fairly stable for the past two years, but can certainly be
> > refined. We'd welcome comments on specific terms or axioms.
>
>Is there some plan to adopt and apply (or at least explicit) lexical and
>structural conventions within SUMO? I am particularly thinking about the
>name of relations with respect to the order of the arguments: for many
>relations (e.g. caseRole relations, and for all functions) the common
>reading convention is followed but for most second-order relations, the
>opposite reading is adopted, e.g. for "instance" and "subclass" whose
>counterparts in RDFS (with the same order of arguments) are named "type"
>and "subClassOf"; the common reading convention (for the same order of
>arguments) would recommend the identifiers "type" and "supertype" (and
>for their inverses: "instance" and "subtype").
>More details and related lexical and structural conventions are in
>http://www.webkb.org/doc/conventions.html
>
>Philippe