Taking a different approach to user modelling and clustering

Abstract

How often have you not been able to retrieve information on the web that is relevant? Most of us find ourselves constantly repeating the same time consuming processes to find a web page which is useful. Through these processes we are exposing a lot of information about ourselves which could be used to create tools that would make relevant web pages easier to find! In the following report a solution id proposed which could aid users in retrieving more relevant search results. An application would collect data and formulate a user model based on a clustering based algorithm which will utilize the queries and browsing history in the user's history. Previous solutions within this field where analysed and an ideal adaptive agent was proposed. By using a prototype of the system an evaluation process takes place. The results revealed that the proposed form of user modelling did in fact yield positive results but further work needs to be done. It was also found that using an auto complete bar to give suggestions is a very effective means of offering assistance. Results also found that using a user model to visualize ones history made retrieving previous search results easier.