on the treatment of certainty (in "Defining N-ary Relations on the Semantic Web: Use With Individuals")

Hello Natasha and all,
My name is Yoshio Fukushige, a W3C fellow from Matsushita Electric
Industrial Co., Ltd.
(Panasonic)
And I'm a member of RDF Data Access Working Group from Matushita

(see http://www.w3.org/People/all#fuku  and
http://lists.w3.org/Archives/Public/public-rdf-dawg/2004AprJun/0138.html
for my introduction)

I'm now working on a presentation of probabilistic knowledge (relation) in
SW,
so I'm very interested in your (including Alan) working group note
"Defining N-ary Relations on the Semantic Web: Use With Individuals"

So please let me present my presnonal comment on the way certainty is
treated in that document.

My comment is on the example 1: "Christine has breast tumor with high
probability"

My concern is about the way of representation of certainty, so it may miss
the main point
of your discussion, in that case, please excuse me.

Your analysis  looks like something like:

"Christine has surely has got a diagnosis which tells that she has  breast
tumor,
and that diagnosis is highly probable."

And you treat the probability as a property attatched to the diagnosis,
an additional information to qualify the (diagnosis) value.

However, what do you think of a framework where
adding a probability is more generalized to adding (modal) information to
a statement.

In a sense, the example statement can be understood as

"It seems with high probability that Christine has breast tumor."

With this interpretation, my analysis is like this:

---from here---
graph A (my analysis)

_:b1
 a :probabilistic_statement ;
 :probability [:grade "high"];
 :consequence _:b2 .

_:b2
 a :statment ;
 rdf:subject :Christine ;
 rdf:predicate :have_disease ;
 rdf:object [a :Breast_Tumor] .

--- till here---

Namely, the probability is connected to the reified triple for the target
statement.

The reason why the probability is not directly attatched to the reified
triple
(like the following graph A) is, to generalize to the case when a condition
is
given
(like the following example B
"If the figure of test b is higher than 100, Chirstine has breast tumor with
high probability"
).

If we are given (additional) condition, or antecedent, it can easily be
added to the upper blank node,
which will represent the rule.

And if additional statement C is made,
"If the figure of test c is lower than 50, Christine has breast tumor with
low probability"
we can attach the combination of the new antecedent and the probability to a
new blank node,
yet sharing the "consequence" (reified triple) node.


My comment may not be a good one, but I hope I can contribute any single
bit.
And it'll be so nice if I can get any feedback from you.

So, thank you for reading it.

Here follows some examples mentioned above.

--- from here---
graph A1 (for comparison)

_:b1
    a :statement :
    rdf:subject :Christine ;
    rdf:predicate :have_disease ;
    rdf:object [a :Breast_Tumor] :
    :probability [:grade "high"] .

--- till here ---

example B
"If the figure of test b is higher than 100, Chirstine has breast tumor with
high probability"

--- form here ---
_:b1
 a :probabilistic_statement ;
 :probability [:grade "high"];
 :consequence _:b2 .
 :condition _b3:

_:b2
 a :statment ;
 rdf:subject :Christine ;
 rdf:predicate :have_disease ;
 rdf:object [a :Breast_Tumor] .

_:b3
    a:    statement ;
    rdf:subject    :figure_of_test_b ;
    rdf:predicate    :higher
    rdf:object "100" .

---till here ---

Exampls C:
"If the figure of test c is lower than 50, Chirstine has breast tumor with
low probability"
is added

--- form here ---
_:b1
 a :probabilistic_statement ;
 :probability [:grade "high"];
 :consequence _:b2 ;
 :condition _:b3 .

_:b2
 a :statment ;
 rdf:subject :Christine ;
 rdf:predicate :have_disease ;
 rdf:object [a :Breast_Tumor] .


_:b3
    a:    statement ;
    rdf:subject    :figure_of_test_b ;
    rdf:predicate    :higher
    rdf:object "100" .

_:b4
    a :probabilistic_statement ;
    :probability [:grade "low"] ;
    :consequence _:b2 ;
    :condition _:b5 .

_:b5
    a:    statement ;
    rdf:subject    :figure_of_test_c ;
    rdf:predicate    :lower
    rdf:object "50" .

---till here ---

Best,
Yoshio Fukushige
 fuku@w3.org

Received on Friday, 23 July 2004 05:41:44 UTC