Hidden relationships discovery through high-level information fusion

We previously proposed an approach and a framework for high level information fusion that we extended in order to manage uncertain information. The fusion algorithm is graph-based and generic and can be parametrized in order to provide different fusion operations. In this paper, we use uncertain graph-based fusion in order to discover social knowledge from a network of information. We aim at discovering hidden relationships between persons within a social network community from a collection of heterogeneous data sets. We apply our approach on data sets provided by the 2014 VAST Challenge. Data is available regarding the working relationships between persons, but the private relationships that may exist between the actors of the scenario are not explicit. Using semantic information fusion, we show that clues about these private relationships can be discovered, and we give information about the certainty associated to the discovered knowledge.