- From: Seth Russell <seth@robustai.net>
- Date: Thu, 16 Aug 2001 11:12:12 -0700
- To: <Patrick.Stickler@nokia.com>, <www-rdf-interest@w3.org>, <scranefield@infoscience.otago.ac.nz>, <sean@mysterylights.com>
From: <Patrick.Stickler@nokia.com> > Then again, there's still something to be said for standards > in the interest of data portability and system interoperability; > and even though I'm less concerned with issues such as a > standard query language for RDF or which ontologies win out > as the most used, etc. I still think that something as fundamental > as resource identity as it relates to serialization is something > that we should be able to trust the standard to define. Sure it would be ~nice~ to have a standard that worked for everybody ... but identifying things is a process not a syntax. We need to start *working* on that process ... and we can't do that without tools. Actually I suspect what will give the best method of identifying things will evolve from a combination of statistical methods, for example Latent Semantic Analysis [1], and group interaction with those automatically generated context nodes to establish agreements based on usage. [1] http://lsa.colorado.edu/ For example we extract all the tems and URLs from all the emails that people who frequent the SemWeb lists have used ... cluster those statistically against the actual documents ... and come out with a navagatable set of terms and documents that people on the groups could collectively mutate (vote for or against by usage). Then assigning URIs to the clusters would not be a problem. ... just a though that has been floating in my mind for some time. Seth Russell
Received on Thursday, 16 August 2001 14:18:52 UTC