On Mon, May 20, 2013 at 7:43 PM, <jmcclure@xxxxxxxxxxxxxx> wrote: as happens often around here, it seems to me we agree more than disagree, but are using different language, but also seem to disagree about what is shiny new and what is old.
William,
Twenty "No Domain" Properties. The examples generally reflect the class of the subject, i.e. the domain of the property. (For example, is a kind of suggests the triple's subject is of class Kind; and is the datatype of suggest the triple's subject is of class Datatype; andis a role in suggests it's of class Role; and so on). This "is a X of" is not more informative than just
(X) of Something where (X) is an instance of class X
Implied is these properties are each without domain. However I wonder about many eg is a datatype for, a property which likely is of domain Datatype; We are really speaking about two different senses of the word 'domain'. I am using 'domain' as in domain of human endeavor, domain knowledge. For example, organic chemistry, baseball, derivatives trading. It seems to me you are using domain in the mathematical sense, in which Datatype is a domain. I guess this is another domain of human endeavor, but a sort of 'meta' domain of endeavor, for computer scientists, not for chemists and ballplayers.
I am asserting that it best to treat everything you want to talk ABOUT as an entity, a thing, is to **cast** it, no matter how it appears in a given discourse, or language, the same way, as a noun. To let this noun represent the concept, instances of the concept, everything you ever want to say about your domain of knowledge. So, catalysis is a noun, catalyst is a noun, atomic weight is a noun, 36.7 is a noun, decimal number is a noun, hydrogen is a noun, water is a noun, etc. etc. ***All** of these represent hunks of something the chemist might be interested in or know about. The role of the grammar and logic of the ontology is to provide the links with which these things can be connected, and so make assertions about their relationships, using non-domain specific concepts like event, role, is a kind of. So, catalyis is a kind of chemical reaction names two domain specific thingies, chemical reactions and catalysis, using one domain independent connector, 'is a kind of'.
is an instance of, is a kind of, is a part of, is a datatype, is a role, is a ternary relationship, these, on the other hand, apply to chemistry as well as to baseball.
and plays the role in, of domain Role), implication being the class of the subject is implicit per the property's name; isn't this under-specification?
Absolutely!!!!
Is it true, for instance, that every conceivable resource plays the role in some-thing either as an actor, a function or otherwise?
CAN play many such roles, depending on role constraints (which often change over time)
That s the very point, it might play the role OF something in something, but the role is a very different kind of entity from an entity that assumes the role. The kinds of roles I am talking about are things like 'brother'. Not things like fido the dog, who might ***play the role** of brother in a sibling relationships with Sallie the dog.
And, I find it incredibly ineffectie to create a hierarchy, in which every role is of 'type' role. I wold rather stereotype some things as roles, or as n-ary relational entities, meaning that they can be combined in certain way with other things, but not treating this combinatorial capability as a kind of class. For example, if R is a role and N is a relation, the it is grammatical to say N has role R. This is just grammar and logic, not content. The content is where all the good stuff happens.
Your goal for properties "devoid of domain specific content" defaults the range of all these properties to rdf:Resource.
But my goal is opposite. There are no properties, except as a gramatical category some things are allowed to play. there is only a logical relationship between two THINGS, like wood and color, and the purely logical relationship that Wood HAS a color. Color is a thing, so is wood. Both can be used as properties, saying this thing is Wood, saying this thing has a color. Or between two other things, like this peice of wood an that particular colur, and the relationship that this peice of wood has that color.
Cannot the model be more specific, relating actors roles events?
Absolutely!!!! That is exactly what a model does. Relates all these kinds of things that have some domain meaning.
I once wrote about the basic model of interest being one representing staged performances, including others like script and scene.
sounds excellent.
"devoid of domain specific content" sounds underspecified to me as above.
I agree with the goal though, to have the range and domain of a property be unbounded. If properties are, say, prepositions & verbs, then they must be without range or domain. Logically the RDF Model suggests doing exactly this, as it defines the term Statement composed of Subject Predicate Object, a very grammatical model.
A very narrow grammatical model, which seems to me to have been created by someone who did not know mathematical logic or much beyond the grammar that used to be taught in schools. I am sorry to hear this about RDF.
Notionally "predicates" decompose to predicate-verb and -object, the latter being a Clause -- a subtype of Statement. Thus properties are named only as verbs & prepositions.
Tenseless Amodal Properties. I'd like to know how you'd model this in RDF triples then
past(Harry believes (John Know Bill))
I know for me it'd be:
Person:Harry had [[Belief:That JohnKnowsBill]]
For me, this would be more like
Past (exists B of type Belief, (Harry of type Person, and Harry plays the role of believer in B, and Proposition:P plays the role of believed in B, and P = (Exists A of type Acquaintanceship such that Person:John plays the role of aquainted in A and Person Bill Plays the role of aquainted with in A).
Without all the subject and predicate stuff, so I would guess less expensive.
Belief:ThatJohnKnowsBill of [[Type:Belief]]
Belief:ThatJohnKnowsBill of [[Type:Clause]]
Belief:ThatJohnKnowsBill subject [[Person:John]]
Belief:ThatJohnKnowsBill predicate [[Property:foaf:knows]]
Belief:ThatJohnKnowsBill object [[Person:Bill]]
I think that in an ontological language, the individual english words are not separate, but part of a
singe meaning, such as 'there exists', and 'for all', and 'if... then'. These just happen to require two words in English.
I think the goals of computational ontologies are to parse model & mine (shake rattle & roll!) those words!
Thanks - jmc
On 19.05.2013 15:58, William Frank wrote:
I left out a few key ones:
is a template for
is a specification of
is in the domain of
is the controller for the domain of
On Sun, May 19, 2013 at 6:50 PM, William Frank <williamf.frank@xxxxxxxxx> wrote:
On Sun, May 19, 2013 at 5:17 PM, <jmcclure@xxxxxxxxxxxxxx> wrote:
My issues with relations named like "is-part-of" are
This is interesting, in that the ONLY relationships I would like to see are of this kind,though 'is-part-of' is the most problematical. So, we have made totally oposite conclusions, based on our mutual decades of practice and study. What does this mean, I wonder.
I have found that the cleanest way to build an ontology is to severely restrict the relationships names, to those that are devoid of domain specific content, using only fifteen, To whit:
is a kind of
is a part of
is a role in
plays the role of in
is an instance of
is a life cycle state of
is a descriptor that applies to
is a data value for the descriptor D
is the data type of
is an actor in
is a policy of an actor in
is a resource used by
is a goal of
is an action of (as many arguments as represent things involved)
- the concept of "part" cannot be qualified without resorting to sub-properties
- the instance of the property itself cannot be qualified, period
Interested to hear what these mean.
- multiple tenses (was-part-of, will-be-part-of, might-be-part-of) balloon all ontologies
I have never heard this view before. What I have always heard, and firmly believe, after decades, is that the essense of an assertion is a proposition that is tenseless and modal less.
so that (John know Bill) shows a picture of the world, and we can modify the proposition with a tense
past(John know Bill)
now (John know Bill)
or modify it with various modalities:
deontic modalities
must (John know Bill)
may not (John know Bill)
epistemic modalities
Harry believes (John Know Bill)
etc.
- the operative term "of" presents itself as a lexical afterthought
I think that in an ontological language, the individual english words are not separate, but part of a singe meaning, such as 'there exists', and 'for all', and 'if... then'. These just happen to require two words in English.
This is only one example. I can think of many others all conforming to verb-noun-preposition used to name a relation between entities (eg is-employed-by).
This would be something many would model as follows.
The impact a property-name exerts across the spectrum -- during input, storage, queries, exchange, etc. -- is one with real-life practical (read, expensive) consequences. Growth becomes unsustainable.
The most widely understood discriminator between "non-semantic" systems and "semantic" systems is that properties are named differently: in "old style" systems these names are nouns, perhaps qualified nouns; in "new style" systems these names are, uh, something other than a noun.
Consequently ontologies I see are near-doubled in size: they have both a class and a property named X, as "old style" names promulgate into "new style" systems.
I argue the industry badly needs consensus about the best practice for how attributes/relations are to be named. It affects EVERYTHING. Ontolog is the only forum I've found appropriate to this question. It really needs to be addressed - so much else is, imho, just a sideshow for all those focused on practical applications of semantic technologies.
Thanks - jmc
_________________________________________________________________
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
--
William Frank
413/376-8167
This email is confidential and proprietary, intended for its addressees only.
It may not be distributed to non-addressees, nor its contents divulged,
without the permission of the sender.
_________________________________________________________________
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 (01)
|