Compound Document Analysis by Fusing Evidence Across Media

In this paper a cross media analysis scheme for the semantic interpretation of compound documents is presented. It is essentially a late-fusion mechanism that operates on top of single-media extractors output and it’s main novelty relies on using the evidence extracted from heterogeneous media sources to perform probabilistic inference on a bayesian network that incorporates knowledge about the domain. Experiments performed on a set of 54 compound documents showed that the proposed scheme is able to exploit the existing cross media relations and achieve performance improvements.