Every information system has, either explicitly or implicitly, an associated theory of information access and a set of assumptions that underlie that theory (in the mathematical or logical sense). A model is an embodiment of this theory. In particular, a retrieval model specifies the details of the document representation, the query representation, and the matching function. The overall objective of the FERMI project is to develop a formal theory of multimedia information retrieval (MIR) and to evaluate it by running experiments on realistic document bases. This theory will be the theoretical foundation of information systems supporting the retrieval of multimedia documents on the basis of their structure and content. A new retrieval model which will embody the above theory is the main aim of the project.
The project will develop a logico-probabilistic retrieval model. Four orthogonal research dimensions will be explored: the logical, the probabilistic, the computational and the multimedia dimensions. At the logical level, the project will indentify a suitable logic, or family of logics, for reasoning on the structure and the content of multimedia documents. The implication relation of the logic would capture the notion of relevance of documents to requests. At the probabilistic level, the project will apply the theory of probability to endow the logic with the treatment of uncertainty. At the computational level, the project will study the computational properties of the logic(s) and uncertainty treatment(s) that are derived at the two other levels, developing, when possible, parallel and probabilistic algorithms to improve the performance of retrieval. At the multimedia level, the nature of different types of multimedia data will be investigated in order to understand the representational and inference mechanisms necessary to perform an effective retrieval on this data. These mechanisms will be then incorporated into the logic. A prototypical implementation of a system representing and retrieving multimedia documents according to the principles of the logic will be developed and used in an experiment aiming at evaluating the adequacy of the logic as a multimedia information retrieval model.
It is expected that the MIR theory will have a significant impact both on the industrial world and the scientific community. As far as the former is concerned, it will allow the development of the technology supporting the construction of a new generation of multimedia document retrieval systems, providing an effective and efficient content-based information retrieval. As far as the latter is concerned, it is believed that the theory will contribute significantly to the definition of a well-founded framework for research in multimedia information retrieval.
The FERMI consortium will exploit the results of the project in several ways. In the medium term, the members of the consortium and industrial groups involved in many industrial projects in the area of multimedia information systems will use specific results of our research as it evolves and is recognised. Therefore, a medium term exploitation will regard the infusion of the project results into several ongoing and starting industrial projects. In the long term, the theory will provide the logical foundation of a new generation of multimedia information retrieval systems that will be required when industry realizes that there are limits to the actual systems. In addition, the consortium will disseminate its research results via the standard academic and research channels. The consortium will also use the IDOMENEUS Network of Excellence industrial liasons and courses to improve technology transfer, particularly to European industry.