- From: Phil Tetlow <philip.tetlow@uk.ibm.com>
- Date: Mon, 11 Oct 2004 11:34:01 +0100
- To: "\"Phil Dawes\" <pdawes@users.sourceforge.net> Patrick Stickler" <pdawes@users.sourceforge.netpdawes>
- Cc: www-rdf-interest@w3.org
Phil I agree with your reasoning, but consider that context does not necessarily require human interrogation to establish credibility. My feeling is that rich stochastic algorithms could be employed over semantic data layers to establish 'degrees' of 'relevance' and 'authority' around any contextural information found (i.e. to what degree can I automatically derive context from the URI under investigation and to what degree can I believe the information that has been presented?). I, hence, personally consider that the trick to success here is establishing and standardising on the algorithms used to automatically derive context and accepting that the answers they return may well not be black or white. This may well have a profound effect on the way that the Semantic Web is perceived and/or used. Rather than using metadata as an ultimate declarative source of descriptive annotation, it may well only represent one piece of a much richer view of the world. For me it is important to remember that all Information Systems, be they automated or not, have there roots in the real world and merely abstract aspects of its the infinite complexity and beauty. Relying on metadata alone to describe such an incalculably encompassing problem space has to be a folly. ‘The answer is that within many applications we should (control capability), but in the Web as a whole we should not. Why? Because when you look at the complexity of the world that the Semantic Web must be able to describe, you realise that it must be possible to use any amount of power as needed. The success of the (current) Web is that hypertext is so flexible a medium that the Web does not constrain the knowledge it tries to represent. The same must be true of the web of meaning. In fact, the web of everything we know and use from day to day is complex; we need a strong language to represent it’ – TBL’s words not mine. Regards Phil Tetlow Senior Consultant IBM Business Consulting Services Mobile. (+44) 7740 923328 "Phil Dawes" <pdawes@users.sou rceforge.net> To Sent by: Patrick Stickler www-rdf-interest- <patrick.stickler@nokia.com> request@w3.org cc www-rdf-interest@w3.org Subject 08/10/2004 18:59 URIQA thwarted by context problems? Hi Patrick, I'm afraid that the more work I do with rdf, the more I'm having problems seeing URIQA working as a mechanism for bootstrapping the semantic web. The main problem I think is that when discovering new information, people are always required to sort out context (a point made by Uche Ogbuji on the rdf-interest list recently). When identifying new terms, some mechanism has to exist to decide whether the author's definition of the term fits with its use in the instance data, and that that tallies with the context in which the system is attempting to use the data. To my mind this prohibits a system 'discovering' a new term without a human vetoing and managing its use. Of course this doesn't prohibit the decentralisation of such context-management work - e.g. a third party could recommend a particular ontological mapping of terms based on an agreed context. I just don't see machines being able to do this work on an ad-hoc basis any time soon. You've been doing a lot of work on trust/context etc.. in addition to URIQA, so I'd be interested to hear your views on this. Many thanks, Phil
Received on Monday, 11 October 2004 10:34:47 UTC