Re: SUO: SUMO questions
Adam,
Sure no problem. I find that It's pretty easy to navigate the ontology with the browser, and in general I find the WordNet mappings pretty intuitive. Thanks for providing the service.
Erik
Adam Pease <adampease@earthlink.net> wrote:
Erik,
Good discussion. Thanks for raising these points. Ian is actually in
the process of creating many new concepts and creating a more specific set
of mappings to those new concepts. So I think b) is a good way of
describing the current approach.
Adam
At 02:52 PM 6/18/2003 -0700, Erik Larson wrote:
>Adam and Ian,
>
>I've been considering this little toy example and think its interesting
>and worth a few more comments.
>
>The problem with finding a noun mapped to a relation only admits of a few
>solutions that I can see, such as a) axiomatizing the meaning (i.e.,
>writing a logic template that gets the intended meaning such as you
>suggested in your prior message) or b) mapping instead to a concept that
>may need to be created or c) just leaving it alone and not considering it
>a !
problem
per se.
>
>In the case of mapping to a concept, if you map to a broader (or "nearby")
>concept then you might lose the essential, intended meaning of the word
>(e.g., a department is ALWAYS a suborganization of some other
>organization). Of course this is the point of mapping to a rule instead
>to get the intended meaning from the ontology without adding new
>vocabulary that may not be suitable for an upper ontology.
>
>So I think its an interesting thing to consider, whether nouns that won't
>map directly to concepts without losing something essential should then
>map to relations (if available), or whether there should be axioms that
>have to explicate the meaning in-terms-of those relations in all these
>cases. As a general solution this would be a lot of work, and anyway most
>people using the browser will just want to get pointed in the right
>direction (in which case
subOrganization would do well enough). I suppose
>this is case c. above.
>
>I think this is a good example of tradeoffs at work in moving between
>words and terms in the ontology.
>
>Anyway I like the quick fix that Ian mentioned which is just to omit the
>confusing example in 3 and include it as an example in 1.
>
>Erik
>
>
>Erik Larson wrote:
>Adam,
>Yes I think you can just map it to Organization. Note however that though
>department is an instance of Organization, it in particular always has
>subOrganization(?DEPT, ?ORG) true of it, which isn't the case for any
>arbitrary instance of an organization.
>
>The "department of literature" as Ian pointed out should be an example for
>sense #1 not #3. I think this is the crux. In combination with this
>example for sense #3, the sense #1 example "specialized division of a
>large organization" steers you away from something like an academic
>department. But in fact I do think that the sense #1 example is really
>similar to the semantics of an academic department. The field of study of
>the group of agents in a given deparment (including the other related
>materials such as publications and so on) make the department a
>"specialized division." The wording is a bit suggestive of corporations
>is all. So perhaps sense #1 suffices modulo considerations of mapping it
>to a relation.
>
>Thanks
>Erik
>
>Adam Pease wrote:
>Erik,
>Many thanks for addressing a substantive technical issue. I see that Ian
>has responded but I'll add my two cents.
>
>There are three WordNet synsets for the English word "department"
>
>1. A specialized division of a large organization; "you'll find it in the
>hardware
department"; "she got a job in the historical section of the
>Treasury".
>
>* SUMO Equivalents: subOrganizations
>
>2. The territorial and administrative division of some countries (such as
>France).
>
>* SUMO Equivalents: GeopoliticalArea
>
>3. A specialized sphere of knowledge; "baking is not my department"; "his
>work established a new department of literature".
>
>* SUMO Equivalents: Proposition
>
>
>
>Each sense maps to a different SUMO concept. So, you are quite right that
>in sense #1 a department has a chair, a location and so forth. However,
>the SUMO t! erm Proposition is mapped to a different sense of "department",
>which is the sense that means "sphere of knowledge". I'm a little unclear
>on sense #3 in that the first sentence makes me think that it is just an
>abstract Proposition, which the second sentence in sense #3 leads me
to
>believe that the mapping should be to FieldOfStudy.
>One could argue about a noun being mapped to a Relation. This is a case
>where the mapping isn't exact because it should really be to a "template"
>logical expressions that addresses the logical meaning of the word in the
>context of a sentence. For example
>
>"Bill is in the hardware department." might be translated to something like
>
>(exists (?P ?D ?O)
>(and
>(instance ?P Male)
>(name "Bill" ?P)
>(location ?P ?D)
>(instance ?O Organization)
>(subOrganization ?D ?O)))
>
>A revised and expanded version of the SUMO-WordNet mappings might have as
>the entry for synset #1
>
>(exists (?D ?O)
>(and
>(instance ?O Organization)
>(subOrganization ?D ?O)))
>
>which would express a fragment of logic that would need to be included in a
>translation to logic of the word sens!
e used in
context.
>One might map to Organization instead of the relation subOrganization
>though.
>I hope this clarifies things somewhat.
>
>Adam
>
>At 10:54 AM 6/18/2003 -0700, Erik Larson wrote:
>
> >Adam,
> >
> > From the SUMO browser for "Department" entered as an English word:
> >
> >A specialized sphere of knowledge; "baking is not my department"; "his
> >work established a new department of literature".
> >
> >SUMO Equivalents: Proposition
> >
> >This can't be right. A proposition is abstract but a department has a
> >chair, a physical address and so on.
> >
> >A department is really an instance of an organization, as the SUMO comment
> >makes clear elsewhere, for Organization. Organizations are
> >groups-objects-physical in SUMO (going up the hierarchy), this is good.
> >
>
>Universities are organizations too, so SUMO's subOrganization predicate
> >(or some specialization of it) can be used to relate departments to their
> >universities. There's SUMO vocabulary for FieldOfStudy (a type of
> >proposition), so all that's required is a binary predicate relating
> >departments with their fields of study. If something like
> >departmentFieldOfStudy is too specific, how about groupContentFocus (ouch)
> >or something similar that relates a group of people to a particular set of
> >propositions constituting the reason for the existence of the group?
> >
> >Erik
> >
> >
> >Do you Yahoo!?
> >SBC
> >Yahoo! DSL - Now only $29.95 per month!
>
>
>Do you Yahoo!?
>SBC
>Yahoo! DSL - Now only $29.95 per
month!
>
>
>Do you Yahoo!?
>SBC
>Yahoo! DSL - Now only $29.95 per month!
Do you Yahoo!?
SBC Yahoo! DSL - Now only $29.95 per month!