Release of the BOEMIE second public annual report

December 28, 2007

Summary of achievements so far: BOEMIE is advocating a novel approach towards the automation of knowledge acquisition from multimedia content. 1. The information extraction toolkit, which implements the BOEMIE extraction methodology as an open architecture. Apart from the development of the methodology, various novel extraction methods for images, video, audio, video OCR and text have been produced and integrated into the first version of the BOEMIE extraction toolkit that has been delivered. Advanced reasoning techniques for multimedia interpretation, based on abductive methods are also included in the extraction toolkit. 2. The ontology evolution toolkit, implementing the BOEMIE evolution methodology as an open architecture. Based on the methodology that was developed in the first year of the project, innovative ontology population, enrichment and coordination techniques have been produced and integrated in the first version of the BOEMIE ontology evolution toolkit, which has also been delivered. 3. The multimedia semantic model, which combines the various ontologies involved in the process, and is augmented with the necessary rules and constraints to form a cohesive and consistent semantic model. In this context, multimedia, geographic and domain ontologies for the domain of athletics have been produced and integrated in the current BOEMIE semantic model. Most of the methods that were developed for multimedia content analysis and ontology evolution in BOEMIE, are trainable, in order to be adaptable to new domains. For the purposes of training and testing these methods, a large corpus of manually annotated content for the various modalities has been produced. For the annotation of this content, new user-friendly annotation tools were developed for textual and visual content. The annotation tools are publicly available, while the public distribution of the manually annotated content for research purposes is currently being negotiated with the owners of the original content. In addition to the work on individual components, the first integrated prototype has been developed and is currently undergoing internal testing. Soon it will be provided for evaluation by external end users of this technology. The feedback that will be received through the evaluation of the prototype will be used for the improved final prototype to be developed in the third year of the project. Furthermore, various advancements of the individual analysis tools and the fusion methods, as well as the acquisition of new concepts and relations are planned for the last year of the project.

download
Tags: