From Observation to Abstraction: A Semantic Web Approach to Convert Sensor Data into Contextual Knowledge and Actionable Intelligence

Abstract

Abstraction is a representation of an environment, derived from sensor observation data. The generation of abstraction requires the ability to infer explanations from an incomplete set of observations (often found on the Web) and to update these explanations based on new information. In addition, this process must be fast and efficient. We present an approach to systematically derive abstractions from observations while overcoming the challenges listed above. In particular, we show how to derive abstractions from observations that conform to a standard sensor ontology, using an abductive logic framework and the inference capabilities of an off-the-shelf OWL-DL reasoner. The utility and scalability of our approach is exemplified through use-cases in the domain of health-care and weather.