- From: Chris Bizer <chris@bizer.de>
- Date: Fri, 19 Dec 2003 12:11:15 +0100
- To: "Jingwei Huang" <jingwei.huang@utoronto.ca>, <www-rdf-interest@w3.org>
- Cc: "Mark S. Fox" <msf@eil.utoronto.ca>
Hi Jingwei, also nice to hear of your work :-) There really seam to be some people spread around the world working on the topic. I think there should be more exchange between the different groups. A thing I try to facilitate a little bit by maintaining the Semantic Web Trust and Security Resource Guide: http://www.wiwiss.fu-berlin.de/suhl/bizer/SWTSGuide/index.htm > We have 4 levels of KP models: Level 1 (Static KP) develops the fundamental > concepts for KP, and focuses on provenance of static and certain > information; Level 2 (Dynamic KP) considers how the validity of information > may change over time; Level 3 (Uncertain KP) considers uncertain truth value > and uncertain trust relations; Level 4 (Judgment-based KP) focuses on social > processes (including indirect trust relations) necessary to support > provenance. > I like your levels. Everything that is structuring the diffuse topic of trust and uncertainness is always useful :-) > > A paper on uncertain KP is in revision and will be available very soon. > Please give me a hint, when it is available. Reading your paper about Dynamic Knowledge Provenance I'm wondering about 2 things: 1. Context and role based trust: Your approach relies on explicit trust statements (as nearly all approaches within the Semantic Web community). In real live may trust decisions are based on context and roles. Like "Trust all doctors in the medical domain" or "Trust people about web programming if they have worked for at least 3 web programming projects". This kind of stuff I tried to capture with my cQL query language. How to you model context and role based trust? Do you use rules for deriving explicit trust statements from context and roles? 2. Scaling of trust: In your trusted_truth_value(a,x,v) statements you are using the trust scaling of "Trusted", "Untrusted" and "Unkown". For many applications I think more fine grained trust scalings are necessary. Lets say you have to choose a subcontractor and you think most of them will do it , but some will deliver better quality. Most theories from social science state that trust should be handled as a continuum. How do you handle the whole topic of information ranking? Is there a general tendency in logic-based approaches to skip information ranking? I think there were some interesting ideas about ranking in [1]. How do Web-of-Trust mechanisms with more fine grained trust stateings fit into your model? Chris [1] http://annotation.semanticweb.org/iswc2003/abstracts/RankingAnnotation.html
Received on Friday, 19 December 2003 06:07:55 UTC