Early detection of significant traumatic events, e.g. a terrorist attack or a ship capsizing, is important to ensure that a prompt emergency response can occur. In the modern world telecommunication systems could play a key role in ensuring a successful emergency response by detecting such incidents through significant changes in calls and access to the networks. In this paper a methodology is illustrated to detect such incidents immediately (with the delay in the order of milliseconds), by processing semantically annotated streams of data in cellular telecommunication systems. In our methodology, live information about the position and status of phones are encoded as RDF streams. We propose an algorithm that processes streams of RDF annotated telecommunication data to detect abnormality. Our approach is exemplified in the context of a passenger cruise ship capsizing. However, the approach is readily translatable to other incidents. Our evaluation results show that with a properly chosen window size, such incidents can be detected efficiently and effectively.