- From: Phil Dawes <pdawes@users.sf.net>
- Date: Mon, 11 Oct 2004 07:56:37 +0000
- To: Giovanni Tummarello <giovanni@wup.it>
- Cc: www-rdf-interest@w3.org
Hi Giovanni, Giovanni Tummarello writes: > Hi Phil, i take the liberty of replying since we've actually studied CBD > like structures (the RDFN in www.dbin.org semantic web p2p) and they are > in fact what makes the whole thing in a sense provingly scalable. > > The point Ogbouji was making was what basically all agree upon, tryng to > do real semantic matching automatically is a lost cause. So .. when one > asks for a CBD discovers "new information" it refers to "new annottions > made using existing, shared ontologies". If the ontologies are unknown > then.. i guess a savy agent will ignore those or ignore the source > alltogether. That sounds reasonable to me. I was specificially concerned about the expectation that semweb agents would be able 'lookup' new terms, rather than just discovering new instance data using well understood and deployed ontologies/schemas. (I probably didn't make that clear). > So, given that there is no intention nor need for uriqua to have to > solve the millenium old problem of AI, I believe the term bootstrapping > is at least in a sense correct and that is what we argue at the > beginning of the paper in the homepage at our site, CBDlike have a very > limited computational cost and can be considered as the "standard > question to ask" the "standard questions that anyone is willing to > answer" (becouse you cant really say "i am open for arbitrary queries" > without opening your computer to easy denial of services). > Sounds interesting - I'll download and read your paper tonight. Many thanks, Phil
Received on Monday, 11 October 2004 09:40:29 UTC