A Representation Language for Semantic Description of Multimedia Contents.
Résumé
This thesis aims at studying the application of the techniques of the semantic Web to multi-media by using shared vocabularies and ontologies. If the low level analysis of multi-media features was largely studied, that of the high level descriptions requires even more effort. Thus this thesis aims at seeking ontology-based description models while founding this one on formal semantics in particular description logics. This would make it possible to precisely translate the expressions of an ontology in another and to carry out logical reasoning on descriptions. Such an approach will allow developing adapted reasoning to carry out classifications, to operate rewritings and optimizations of queries requests, and to infer approximate answers.
The standard aspect of the semantic languages coming from the fact that they are conceived for all types of resource leads us to work for specific extensions to multimedia. Then and during my thesis I will be concerned about : On the one hand conceiving a language and establishing an ontology of the field allowing for specific multimedia description including a symbolic system part, a visual part and a common part binding the top and of low level representations. On the other hand developing adapted reasoning and detectors allowing for an automatic extraction of high level characteristics based on the preceding language in order thus to contribute in the reduction of semantic gap.