Design and Implementation of a Cross-Media Indexing System for the Reveal-This System

Despite the vast growth of heterogeneous, multimedia and increasingly multi-lingual digital content, there is a lack of integrated technology that facilitates its effective usage. The need is being expressed insistently by end-users, and professionals in content business, particularly, with respect to the re-use and re-purposing of multimedia content. The EU-IST Reveal-This (R-T) project addresses this fundamental need aiming at developing an umbrella technology able to provide retrieval, categorisation, summarisation and translation functionalities that will help users keep up with the explosion of digital content. The system is being developed to cope with digital content that is scattered over different platforms, different languages and different types of media. In order to effectively achieve this vision, the project sets the following two key technical and scientific objectives: 1) benefit from cross-media analysis and indexing techniques and 2) develop representations suitable for cross-media and language technologies that can extend the potential of content push and pull technologies. Research carried out in the project attempts to address all these issues. Inline with the objectives and central to the project, we propose an architectural unit called Cross-Media Indexing Component (CMIC) which constitutes the final layer of the Cross-Media Analysis and Indexing Subsystem. It is employed to cross analyse and leverage the individual potential of each indexing information generated by the analyzers of diverse modalities. In this paper, we focus on CMIC's role and its support for retrieval. The indexing and evaluation approaches are presented together with the status of work in progress.