Architecture for semi-automatic multimedia analysis by hypothesis reinforcement

The digitalization of the audiovisual production chain has introduced new opportunities and challenges in the asset management workflow. The huge amount of accesible content requieres new annotation and indexing paradigms that overcome the current limitations in terms of resources and level of detail. A novel approach to improve and automatize professional Media Asset Management systems is proposed in this paper. Our proposed architecture enhances the metadata with new objective concepts that can be ported to the semantic level and can also used through ldquoquery by samplerdquo methods. Moreover, the implicit and explicit knowledge about a certain domain can be introduced in the system with a combination of classifiers and a semantic middleware. Last, the system can be replicated in different domains and combined via an initial hypothesis, allowing the scalability of the system to multiple content domains.