- From: Thomas B. Passin <tpassin@comcast.net>
- Date: Mon, 20 Sep 2004 17:59:26 -0400
- To: www-rdf-interest@w3.org
Harry Halpin wrote: > The very idea of a "referent" or a "thing" > is a bit problematic - while I agree an RDF predicate denoting that would > be good, we're going to go right up against one of the largest problems > with all knowledge representation systems, SW included - how to know > if what I mean by my "thing" is the same as what you mean? No matter what, this problem - what does an identifier identify, and is it the same as that identified by another identifier - is always going to be with us, because it has no general solution. The best we can hope for is to have some techniques available that help in some number of practical cases. In some cases, we can use owl to say that two individuals or two properties are the same (if we happen to know that). In some cases, inverseFunctional properties will allow a processor to know. In other cases, we may have to go with probability, when enough nonInverseFunctional properties exist (if my wife, address, height, hair color, apparent age, etc., etc., meet the description, then I have probably been identified). Maybe with Web Proper Names, we can get some more help by means of parsing the URI (or whatever representation). It seems to me to be basically a vehicle for packaging up a number of non-inverseFunctional properties in a compact way, judging by the few minutes I have been able to spend looking at the paper so far. That could be useful, I would think. Cheers, Tom P -- Thomas B. Passin Explorer's Guide to the Semantic Web (Manning Books) http://www.manning.com/catalog/view.php?book=passin
Received on Monday, 20 September 2004 21:56:28 UTC