Real-time task recognition based on knowledge workers' computer activities

Motivation -- Supporting knowledge workers in their self-management by providing them overviews of performed tasks. Research approach -- Computer interaction data of knowledge workers was logged during their work. For each user different classifiers were trained and compared on their performance on recognizing 12 specified tasks. Findings/Design -- After only a few hours of training data reasonable classification accuracy can be achieved. There was not one classifier that suited all users best. Take away message -- Task recognition based on knowledge workers' computer activities is feasible with little training, although personalization is an important issue.