We want to create a basic system that allows the unsupervised extraction of consistent and meaningful semantic information. Central goal of the DIRECT-INFO system is to offer an integrated system combining the output of basic media analysis modules to semantically meaningful trend analysis results which shall give executive managers, decision and policy makers a solid informational basis for their strategic decisions. DIRECT-INFO will address concrete usage scenarios that are common to virtually all advertising and media monitoring industries. However, the system will be designed in such a way that it is open for covering further scenarios in additional sectors such as government, policy/PR analysis and decision making, financial (decision making, trend analysis) and other corporate and media scenarios. Hence the system shall provide an innovative solution in the area of decision support systems. It will significantly ease necessary information filtering and analysis steps that are important in decision-making processes.
Technically the system shall integrate the information of audio/visual, text, audio and video data with a comprehensive, end-to-end solution for the collection, identification and, analysis of media information. It shall utilise the multi-modality and event-driven characteristic of multimedia content for a wide range of applications as for example the extraction and annotation of stories from broadcasts, the detection of highlights and summarisation of sport events or monitoring solutions. DIRECT-INFO therefore will apply a new approach for indexing of multimedia data, which takes advantage of the media's multi-modality.
The goal is to first detect and extract logical entities (context) from the data by pre-specified schemes for different kind of information. In the second step, the extracted entities will then be semantically correlated by taking advantage of predefined and learned contextual knowledge and semantic information.
Funding SchemeSTREP - Specific Targeted Research Project