The advent of sensor technologies and the Semantic Web provide the unique opportunity to unify the real and the virtual worlds. It enables the building of very large infrastructures which for the first time facilitate the information-driven real-time integration of the physical world and computers on a global scale. Flexible middleware technology which abstracts from heterogeneous sensor network technologies to higher-level functionalities to enable interconnected sensor networks and processing of sensor data (Sensor Internet) is a cornerstone of enabling Semantic Reality. The sheer size of this system poses quite novel and unique challenges, as it can only be engineered and deployed if a large degree of self-organisation and automation capabilities are being built into the system and its constituents, enabling simple deployment (plug-and-play), dynamic (re-) configuration, flexible component and information integration, and tailored information delivery based on user context and needs in a service-oriented way. This requires semantic descriptions of the user needs and contexts, and of the system's constituents, the data streams they produce, their functionalities and their requirements to enable a machine-understandable information space of real-world entities and their dynamic communication processes on a scale which is beyond the current size of the Internet.