Springer Science and Business Media LLC

Personalised information systems require a profile of the user in order to function effectively. Typically such systems require users to supply a set of keywords which describe their interests. The research presented in this paper derives user interest profiles automatically by monitoring user web and e-mail habits. A clustering algorithm is employed to identify interests, which are then clustered together to form interest themes. User profiles must also adapt to the changing interests of the user over time. This paper shows that interest themes can be tracked over time through measuring the similarity of interest themes across time periods.