This document discusses three ways in which a knowledge model can be underspecified:
1) Using literals instead of things to model concepts like hobbies, which prevents relationships between concepts.
2) Modeling authorship and books without using reification, which limits adding details like author rank.
3) Using a basic friendship ontology that does not distinguish friend levels, preventing labeling some friends as "close." In each case, extending the ontology resolves the underspecification.
3. Literal vs Thing
Suppose I want to model my hobbies, say
Programming:
1st approach:
2nd approach:
3
4. Literal vs Thing
So far so good, but what if I want to model that Programming
is actually a kind of Sports? (*)
Then, the former approach can’t do the trick (because it’s a
literal and a literal can’t be the subject of a triple, at least in
standard RDF).
The latter, fortunately, can be used to model it, by saying:
Thus, I’d say that the 1st approach is underspecified with respect to
the requirement (*).
4
9. Reification
So far so good, but what if I want to model the
ranks of authors, say, my dad is the 1st author,
my mom is the 2nd author and I, myself, is the
last author of the book? (**)
Well, you can’t always get what you want, or..
9
14. Weak Ontology
Suppose that my initial ontology is only like this:
such that it is used to say the friendships of
people.
14
15. Weak Ontology
Suppose I wanna say that, Justin Bieber is my
friend:
Good so far. But, what if I want to say, JB is not only
my friend, instead, he is my close friend! (***)
Well, you can’t do that, unless..
15
16. Weak Ontology
I edit my ontology to become like:
Thus, I can happily model my friendship with JB as
follows:
16
17. Weak Ontology
Thus, I’d say that my initial ontology is
underspecified regarding the requirement
(***).
17