Simon,
The linguistics presentation is interesting, but the issue I was addressing was confusing linguistics with ontology development.
The linguistic expectation is that “pineapple tree” refers to a kind of “tree” on which pineapple fruit grows, by following exactly the interpretation conventions you identify. The presumption that the reference of “tree” in this compound is a plant is based on the observation that the specializer “pineapple” refers to a fruit. That is, ideas 2 and 3 are used to select the interpretation of “tree” that is being specialized. That assumption would not be made for “shoe tree” or “family tree” or “syntax tree”. But it may be possible that someone’s term for the pineapple plant is “pineapple tree”, in yet another extension of the term “tree”. For example, fruit growers use “tree” to refer to the woody plant on which the fruit grows, as distinct from the fruit itself, even when the plant is technically a shrub. This is why it takes something like Watson’s massive knowledge base AND Bayesian reasoning to determine what is (probably) meant by natural language.
The whole idea in ontology development is that we sort all that out in the formal language. In a formal ontology, all terms are Humpty Dumpty words: They mean exactly what we say they mean, neither more nor less.
This leads to a related observation: An ontology in which most of the terms are primitive (not formally defined) does not provide much of a foundation for inference. In particular, most OWL models I see are information models, not ontologies. They don’t DEFINE terms using both necessary and sufficient (iff) characteristics. They specify only the necessary and possible characteristics of a type, as one does in ORM or UML. And in at least two “professional examples”, the problem is a kind of illiteracy: The modeler states that A is a subtype of B and C, when s/he means that A is the intersection of B and C. Or similarly that A is a subtype of B and has property P, when what is meant is that a thing is an instance of A exactly when it is a B and has property P.
(From the point of view of socializing modeling, I am happy to see OWL used as an information modeling language by people who would otherwise draw Powerpoint diagrams or use XML Schema to capture their domain models. But that only means we have finally rescued conceptual modeling from the captivity of database design and object programming. As I have said before, the use of OWL (or UML or ORM) to capture conceptual models, as opposed to software designs, is a major step forward in knowledge engineering. The problem we now face is not to confuse using OWL with capturing a real ontology in OWL. Put another way, I think UML models written in Turtle or Manchester Syntax are good; calling them “ontologies” is not good.)
-Ed
--
Edward J. Barkmeyer Email: edbark@xxxxxxxx
National Institute of Standards & Technology
Systems Integration Division
100 Bureau Drive, Stop 8265 Work: +1 301-975-3528
Gaithersburg, MD 20899-8265 Mobile: +1 240-672-5800
"The opinions expressed above do not reflect consensus of NIST,
and have not been reviewed by any Government authority."
From: ontolog-forum-bounces@xxxxxxxxxxxxxxxx [mailto:ontolog-forum-bounces@xxxxxxxxxxxxxxxx] On Behalf Of Simon Spero
Sent: Tuesday, April 02, 2013 6:23 PM
To: [ontolog-forum]
Subject: [ontolog-forum] Compound nouns
On Tue, Apr 2, 2013 at 2:11 PM, Barkmeyer, Edward J <edward.barkmeyer@xxxxxxxx> wrote:
There is nothing inaccurate about having the term "pineapple tree" in your ontology. The inaccuracy would be in saying that it is a subtype of "tree", assuming that we all agree on the definition of "tree". The assumption that the word "tree" appearing in the term "pineapple tree" implies some well-defined relationship is unwarranted. This is the kind of thing that comes from trying to guess what is meant by looking at natural language cues. Sometimes you guess wrong!
Fabian Neuhaus's favorite example is "fake diamond". It is by definition NOT a kind of "diamond".
The meaning of the phrase pineapple tree is endocentric - that is to say, the meaning of the whole is derived from its constituent parts.
It is is about as close to canonical an English noun-noun compound as they get (which isn't very).
1. It is right headed (tree is the head noun, and pineapple is the modifier).
2. The specialization in meaning obtained via the modifier is the prototypical specialization for other modifiers in the same general category (an apple tree is a tree on which apples grow; a cherry tree is a tree on which cherries grow.
3. If pineapples grew on trees, they would grow on pineapple trees.
There is considerable difference between pineapple tree and syntax tree ; one would not expect to be able to build a tree house in the latter, nor to hang a bird house from its branches.
Fake diamond is much more complicated to analyze. Syntactically, diamond is still the head, but semantically, modifier fake carries with it the partial meaning that it is something that has a very strong superficial resemblance to a real diamond, but which lacks some critical property. The common super-ordinate category would appear to be things which resemble diamonds.
More complicated still is the case of toys. A toy dog is a toy shaped like a dog, but a toy poodle is a dog (and a dog toy is a toy for use by dogs).
For more fun, consider White tigers, Paper tigers, Paper Airplanes, and Model airplanes.