A business-driven framework for automatic information extraction in professional media production

This paper aims to propose a service-oriented framework for performing content annotation and search, adapted to the task context. Media production workflows are becoming increasingly distributed and heterogeneous.The tasks of professionals in media production can be supported by automatic content analysis and search and retrieval services.

The processes of the framework are derived top-down, starting from business goals and scenarios in audiovisual media production. Formal models of tasks in the production workflow are defined, and business processes are derived from the task models. A software framework enabling the orchestrated execution of these models is developed.

This paper presents a framework that implements the proposed approach called Metadata Production Management Framework (MPMF). The authors show how a media production workflow for a real-world scenario is implemented using the MPMF.

The authors have demonstrated the feasibility of a model-based approach for media processing. In the reification step, there is still information that needs to be provided in addition to the task models to obtain executable processes. Future research should target the further automation of this process.

By means of this approach, the implementation of the business process defines the workflow, whereas the services that are actually used are defined by the configuration. Thus, the processes are stable and, at the same time, the services can be managed very flexibly. If necessary, service implementations can also be completely replaced by others without changing the business process implementation.

The authors introduce a model-based approach to media processing and define a reification process from business-driven task models to executable workflows. This enables a more task-oriented design of media processing workflows and adaptive use of automatic information extraction tools.