Tech Reports

Tech Report kmi-04-02 Abstract

Nootropia: a Self-Organising Agent for Adaptive Document Filtering
Techreport ID: kmi-04-02
Date: 2004
Author(s): Nikolaos Nanas, Victoria Uren, Anne de Roeck, John Domingue

This paper presents Nootropia, a self-organising information agent, capable of evaluating documents according to a user's multiple and changing interests. In Nootropia, a hierarchical term network that takes into account term dependencies is used to represent a user's multiple topics of interest. Non-linear document evaluation is established on that network based on a directed spreading activation model. We then introduce a process for adjusting the network in response to changes in user feedback. We argue that Nootropia exhibits self-organising characteristics, which, as demonstrated experimentally, allow Nootropia to adapt to a variety of simulated interest changes.