- From: Bill dehOra <BdehOra@interx.com>
- Date: Fri, 19 Jan 2001 09:45:47 -0000
- To: "'Sean B. Palmer'" <sean@mysterylights.com>, Bill dehOra <BdehOra@interx.com>, "'Dan Brickley'" <danbri@w3.org>, "McBride, Brian" <bwm@hplb.hpl.hp.com>
- Cc: Tim Berners-Lee <timbl@w3.org>, www-rdf-logic@w3.org
:> Suppose we wanted to do this for real. 'Fred' is not a URI. : :*sigh* This is the age-old problem where we want to make :assertions abotut :things that don't have URIs in RDF. My take on this is that you simply :*give* them a URI:- : :@prefix : <#> ::bill :loves :jane : :Now you don't need to explicitly have a URI "<#bill>" for :<#bill> to exist: :you just talk about it, and call it a person as best you can:- One cautious, baseline way of thinking about string literals, is that for machine purposes they are non-unique and have no semantics. At some point we have to jump from machine meaningless to machine meaningful. If this is the case, then what does making resources out of literals give us? There is at least one sensible answer: allow them to become the subject of statements, since, what we really want to do with a string literal is predicate it as soon as possible. Naturally we want to do this in an automated way. I emphasise machinery here, since it is an easy trap to read semantics into things which aren't there for the purposes of computation. Human readable formats like XML do have a downside. In this case it's largely irrelevant that they look like names to us. Now one way I can see of lifting a literal into a name is having some semantics around the predicate "loves". So a machine processing this: bill :loves :jane can infer a hypothesis that the subject and object of this statement are proper names/nouns. That's a narrow range for "loves" though. Consider, the statement (which is more likely to be true in any case): bill :loves :scotch So given that predicates such as loves will impose ambiguity, we'll need software that can generate an inference to the best possible hypothesis of the subject and object of any statement that in turn has the predicate loves (where ambuigity really means more than one hypothesis can fit the data). In essence, when I need to deal with string literals in RDF, I want my machine to put on its natural language processing/tagging hat. And hope it has a good abductive inferencing engine :) I also note that conceptual graphs is a particulary useful technology for this sort of problem. :Do not start writing things about string literals, :because it doesn't mean anything: write about URIs, please :-) Well we have to articulate something about them, they're a normative aspect of the M&S after all. -Bill ---- Bill de hÓra : InterX : bdehora@interx.com
Received on Friday, 19 January 2001 04:47:34 UTC