In the 20th century, the progress of telecommunication was characterized by the ambition to extend the functionality of monolithic communication systems. The participants were reachable over a few, but well defined, communication channels. The technological improvements in the early 21st century introduce a huge number of parallel available and heterogeneous communication means, including fixed line telephony, mobile telephony, SMS and E-mail. Still the efforts to successfully establish a call increase due to the fact that he has to determine a convenient and available communication mean. The addressee of a call may be disturbed by calls in inappropriate situations. Nowadays, new technologies in the environment of users provide an increasing amount digitally available information, which can be used to describe the situation of a user (e.g., an address book or a calendar on a mobile phone, positioning systems, home automation, …). A mechanism to evaluate this amount of information, i.e. the determination of the context of a user, can be used to support communication services in order to determine an appropriate communication mean. The decision which form of communication is chosen depends on the situation, the relation between the participants, as well as to the availability and the abilities of the user’s devices. The basic challenges, which are addressed by this work, are the provisioning, the determination, the retrieval and the processing of relevant and expressive information sources in a way that the resulting set of information provides a sufficient basis for a subsequent decision-making algorithm. Moreover the work uses semantic descriptions about the abilities and relations of sensors, in order to determine the relevance of information sources. The subsequent evaluation of the information is adaptive to a dynamic set of information, as well as to the behaviour and wishes of the participants. A central aspect of this work is the adaptive evaluation of the information, dealing with highly dynamic and heterogeneous information. This work specifies the requirements, the architecture, and the development of a context service. This context service provides the generation of an evaluation model for context determination by monitoring the user's behaviour in dependency of the retrieved information. Existing algorithms for model generation are analysed, compared and optimized for the targeted application scenario. A proof-of-concept confirms the feasibility of the developed system, and an analysis of the developed algorithms shows a significant improvement of both quality and efficiency of the context determination when compared to existing approaches. | English |