- From: Bijan Parsia <bparsia@cs.manchester.ac.uk>
- Date: Wed, 25 Mar 2009 20:01:56 +0000
- To: "Matthias Samwald" <samwald@gmx.at>
- Cc: "Oliver Ruebenacker" <curoli@gmail.com>, "Mark Wilkinson" <markw@illuminae.com>, "Phillip Lord" <phillip.lord@newcastle.ac.uk>, "W3C HCLSIG hcls" <public-semweb-lifesci@w3.org>
So, in numerics it's pretty easy (well, known, at least). Just trying the pulling out of my ass and hoping for rainbows trick, I can think of a few different "weaker" relations: 1) Configurable identity: You want the objects to be treated as the same for some purposes and (potentially) different for others. In numerics pointer identity vs. coerced equality. For example, in the part of your ontology where you are reasoning about the relation between records and genes you want them to be distinct, but in other places you want to conflate them. I have no idea how to do this :) 2) Similarity: There are description logics based on metric spaces which allow you to reason about the distance between two objects in terms of the number of shared features. 3) Substitutability: You might be able to combine the above two notions. 2 is the best understood and is probably otherwise useful. We have the logics. They seem reasonably implementable. Cheers, Bijan.
Received on Wednesday, 25 March 2009 20:02:32 UTC