Re: Another comment about confidence value.

I would suggest to consider confidence factors as probabilities 
associated to assertions (like "this test has failed").

In this way they have a very clear semantics (by the way, this is what 
confidence factors most often meant in artificial intelligence research 
done on expert systems some years ago). This will pay back in several 
ways, for example if somebody decides to run some experiment to find 
those numbers in some specific report. Or if somebody wants to use those 
numbers to produce a summarization, like "what is the joint probability 
of test results for tests implementing checkpoint 1.1").

Then, depending on specific needs, the way in which a probability level 
is described may be through a numeric scale (eg. p \in [0,1]) or through 
an ordinal scale like ("low", med", "high").

Best regards,
-- 
         Giorgio Brajnik
______________________________________________________________________
Dip. di Matematica e Informatica   | voice: +39 (0432) 55.8445
Università di Udine                | fax:   +39 (0432) 55.8499
Via delle Scienze, 206             | email: giorgio@dimi.uniud.it
Loc. Rizzi -- 33100 Udine -- ITALY | http://www.dimi.uniud.it/giorgio


Charles McCathieNevile wrote:
> 
> On Mon, 18 Apr 2005 13:42:45 +0200, Shadi Abou-Zahra <shadi@w3.org> wrote:
> 
>>> This is the basis of my interoperability concern, and the reason
>>> I wondered if we wanted it at all. Whatever range of values we
>>> pick isn't nearly as important as the actual process used to
>>> assign a confidence rating to a particular result. So I would
>>> like to model the result as a blank node, and the confidence as
>>> a datatype.
>>
>>
>>
>> Yes, the confidence property seems to me to be very important too. And 
>> I  agree that the process model of assigning the value is probably 
>> even  more important than the value itself. However, it is a big 
>> concern to me  if we do not a define a datatype. At the most, may be a 
>> couple of values  with some sort of conversion scheme between them but 
>> I think we are  going to get really big interoperability problems if 
>> we do not define  values.
> 
> 
> I think we are as likely to get interoperability problems by definng  
> smethng as by not doing so - especially if we don't leave it optional.
> 
> But we do clearly need to explain how to define one - which among other  
> things means reviewing the work coming out of the Semantic Web Best  
> Practices group on how to define a datatype (that was left as a work 
> item  by the RDF core group, although the relevant task force is I think 
> at the  point of publishing a draft).
> 
> More when I am not rushing for another plane...
> 
> cheers
> 
> Chaals
>

Received on Monday, 18 April 2005 14:33:39 UTC