Context-driven data and information fusion

One of the major trends of modern data and information fusion technology is to take into account not only the fact of multiple sources but also the available context of a domain. An example case of the above statement is when data to be fused are represented in an object database and can be enriched by an expert with the domain ontology which provides a domain context to each instance of the data. The paper proposes context-driven data and information fusion technology. The latter comprises ontology-driven generation and aggregation of features as well as a two-step filtering of features: based on discriminative power at the first step and on a measure of causality at the second one. A peculiarity of the resulting solution is that every class of decision is specified by a specific set of features represented in terms of predicates that are statements about feature properties. The technology is implemented and validated in several applications of real life scale. In the paper, the technology is demonstrated by the example of applying it to the implementation of personalized intelligent MS Outlook e-mail assistant.