The University of Edinburgh -
Division of Informatics
Forrest Hill & 80 South Bridge

MSc Thesis #9582

Title:"Alexandros" Interface Agents in Self-Organizing Systems: a Prototype
Date: 1995
Abstract:This thesis presents a prototype implementation of a personalized information filtering system to be used in Internet Browsers (e.g., Mosaic, Netscape, etc). In this phase it has been tested and implemented in a browser-like sample database. A personalized information filtering system must be able to provide specialization to user interests but also must be flexible enough to be able to adapt to interest changes. In order to achieve the optimum adaptation level, the system uses interface agents, i.e., autonomous programs that learn the user's interests and act accordingly. The learning mechanism in this phase is based mainly on the genetic algorithm whereas relevance feedback is of minimal use. Another issue that is addressed in this work is self-organization. The ultimate goal of a self-organizing system is to be fully independent. The aspect that is examined in this thesis is the integration of the environment. Experimental results prove that "Alexandros" is able to cope with both adaptation to user interests and integration of the environment. Moreover, there are several suggestions for further development.

[Search These Pages] [DAI Home Page] [Comment]