Re: any good hacks for these?

On Thu, 2002-08-01 at 17:27, Danny Ayers wrote:
> 1. A context thing. Say we have an RDF file containing a set of assertions,
> and we want to load this into our system. The only thing is, we need to know
> that these assertions came from the same (identified) source. Internally we
> can use a quad type model or whatever, but if we want to merge these
> assertions with another set and pass them on, how can this be done whilst
> maintaining an association between the source and the assertions?
> ...
> 3. Weightings/confidence factors. I'll probably (0.5) be having cocoa
> shortly. Or even, isEquivalentToMaybe[-.1].  Any neat approaches?

For what it's worth, Marina Zhurakhinskaya's Master's Thesis addresses
some similar problems in a fairly straight-forward manner.

http://haystack.lcs.mit.edu/papers/belief.pdf

Far from helping, I'd like to add a fifth problem that has been
troubling me lately.

5. Defeasible rules.  It would be great to say in CWM or some other
reasoning engine that people have two biological parents _unless
explicitly declared otherwise_, so I can handle cases like the Biblical
Adam and future cloned children.  Note that given a good standard
solution for the statement "The current context does not contain a
statement {Gen:Adam par:hasNonstandardParentRelationship []}", I could
fake this, although it would be even better to be able to not have to
mention the exceptions in the rule.

Any ideas on this one?  Thanks, all.

	David Saff

Received on Friday, 2 August 2002 11:46:40 UTC