Automatic derivation of context descriptions

Context-awareness in mobile information systems bears a huge potential. However, context-awareness is still in its infancy and its full potential is not yet exploited. One reason is the poorly supported creation and learning of suitable context descriptions. Another problem is the questionable predictive power of context descriptions that makes it difficult to correctly determine the current user context. For applications that depend on the user context, the reliable determination of the context is essential. In this paper, we propose a process to characterize contexts. We correlate raw contextual information with user activities to determine accurate context descriptions. In a case study, we show how different statistical methods can be used to determine correlations, and analyze their applicability.