RAPS: A Topic- and Rank-based Publish / Subscribe System

28.10.2005

Abstract
In the tightly connected information-rich modern world, many bits of information compete for our attention, and finding the relevant bits is increasingly challenging. A common approach to this information overload is to delegate the task of finding the relevant information to someone else through the use of subscriptions. This have led to the emergence of a growing number of publish/subscribe systems, that pushes information to subscribers based on subscriptions. A problem with conventional publish/subscribe systems, however, is that they offer no mechanisms for controlling how much information is received. This means that the user may still be overwhelmed by being pushed to much.

We argue that information overload can be prevented by ranking information based on quality and relevance, and sending the highest ranking information to the users up to the threshold they specify. In this thesis we present RAPS, a topic-based publish/subscribe system that allows subscribers to send explicit feedback - in the form of numeric ratings - on the information they receive. This feedback can be used by publishers to compute collaborative ratings that are used in republishing of information. This allows for swift quality assessment of information, and makes it possible for subscribers to filter information based on how it is ranked. Through an experiment involving users we show that user feedback is an effective way of ranking information.