Re: Certainty Factors

Danny Ayers wrote:
> 
> I'm just wondering what angle the proposed RDF spec takes for dealing
> with uncertainty, and the 'truth' of a statement.

The RDF does state that all statements is considered facts. It says
that there can be a reification of a statement whithout the statment
itself.  This means that you can describe that someone states what
they believe to be a fact, without yourself having to consider it a
fact.

You could in theory import a large body of statements from a foreign
model, as reified statements only. You say that all these statements
is the opinion of that particular source.

You could, after your own reasoning, promote a select few of the
statements to be facts even from your own perspective.


In this sence, there is a binary state of trueness in the model.  I
have choosen to implement this as a bolean value coupled to every
statement in my RDF database.



I think that it would be good to have a common way to state the
Certainty of a specific statement. And I think that it would still be
practical to consider some things absolute facts. 

Given this, you could easily expand the system by asigning a Certainty
to statements. But the statement would only be promoted to fact by the
reasoning of the application. One of the application rules could be to
promte statements with a Certainty of .95 or higher, there the
Certainty statement belongs to the local model.


The Certainty of a statement could be derived by the degree of trust
for the origin of the model of the statement. exampel: A foreign model
could set the Certainty for a statement to .50. Your general trust of
the facts in models from that source could be .10. You could then add
a new Certainty to the statement with the value of .05 and setting
this to be a local statement.


For every operation of the application, you would have to think about
if you are intrested only in facts or also in probabilities.

-- 
/ Jonas  -  http://paranormal.o.se/perl/proj/rdf/schema_editor/

Received on Monday, 13 December 1999 16:35:31 UTC