Classification of user tasks by the user behavior

On-line newspapers have become an important source of information for many people. Readers of on-line newspapers may regularly check the news in general, they may follow the news concerning a certain topic, e. g. when people wanted to get all information that was available on September 11, or they may just check for certain facts such as stock prices or the weather forecast. This thesis deals with the question of whether it is possible to recognize the kind of task a Web user is performing by just looking at the user's behavior. Two studies were conducted to examine whether behavioral aspects, such as mouse and scroll movement, mouse clicks or page view duration, give hint on the surfing mode which may be just browsing, information gathering or fact finding. This thesis is meant to provide a basis for the personalization of Websites such as on-line newspapers where personalization is not only based on the analysis of content information or general topic preferences, but, additionally, on the current context represented by the kind of task a user is performing at the very moment.