Information Extraction from Multimedia Documents for e-Government Applications

Despite the exponential growth of information systems for supporting public administration requirements, we are still far from a complete automatic e-government system. In particular, there exists the need of automatic or semiautomatic procedures for the whole flow of digital documents management, in particular regarding: (1) automatic information extraction from digital documents; (2) semantic interpretation (3) storing; (4) long term preservation and (5) retrieval of the extracted information. In addition, in the last few years the textual information has been enriched with multimedia data, having heterogeneous formats and semantics. In this framework, it's the author's opinion that an effective E-Government information system should provide tools and techniques for multimedia information, in order to manage both the multimedia content of a bureaucratic document and the presentation constraints that are usually associated to such document management systems. In this paper, we will describe a novel system that exploits both textual and image processing techniques, in order to automatically infer knowledge from multimedia data, thus simplifying the indexing and retrieval tasks. A prototypal version of the system has been developed and some preliminary experimental results have been carried out, demonstrating the efficacy in real application contexts.