NOVELTY - Method consists in identifying user characteristics to create a user model, receiving event (e-mail message, task scheduling, information broadcast) data, inputting the data to rules and processing the rules to generate an output representing user action, such as programming a device to record broadcast information. The outputs are combined to generate a predicted action associated with the user which is compared with observed user action to modify the user model, which comprises probabilities or their intervals associated with user characteristics. USE - Method is for e.g. identifying and conveying e.g. TV programs and music to a user, classifying car sportiness etc. ADVANTAGE - Method enables prediction of user actions.
|Patent number||WO2003012704-A; EP1280080-A1; WO2003012704-A1|
|Publication status||Published - 2003|