Journal of Logic and Computation Advance Access originally published online on September 30, 2008
Journal of Logic and Computation 2009 19(5):771-790; doi:10.1093/logcom/exn045
| ||||||||||||||||||||||||||||||||||||||||||||||||||
This article appears in the following Journal of Logic and Computation issue: Special Issue: Recent Advances in Ontology Dynamics [View the issue table of contents]
Original Articles |
Trust-based Revision for Expressive Web Syndication
College of Information Studies, University of Maryland, College Park, MD 20742, USA.
E-mail: jgolbeck{at}umd.edu
Clados Management LLC, San Mateo, CA 94404, USA.
E-mail: christian{at}clados.com
Received 7 November 2007.
 |
Abstract |
|---|
Interest in web-based syndication systems has been growing as information streams onto the web at an increasing rate. Technologies, like the standard Semantic Web languages RDF and OWL, make it possible to create expressive representations of the content of publications and subscriptions in a syndication framework. Because these languages are based in description logics, this representation allows the application to reasoning to make more precise matching of user interests with published information. A challenge to this approach is that the consistency of the underlying knowledge base must be maintained for these techniques to work. With the frequent addition of information from new publications, it is likely that inconsistencies will arise. There are many potential mechanisms for choosing which inconsistent information to discard from the KB to regain consistency; in the case of news syndication, we argue keeping the most trusted information is important for generating the most valuable matches. Thus, in this article, we present algorithms for belief-base revision, and specifically look at the user's trust in the information sources as a metric for deciding what to keep in the KB and what to remove.