Fusion de multi-modalités et réduction par sémantique latente

This paper studies Latent Semantic Analysis (LSA) between visual and textual elements of a multimedia collection, applied on two different tasks: (1) multimedia document retrieval (MDR) and (2) automatic image annotation (AIA). Latent semantics, usually used for text indexing, is applied to discover hidden semantic relations between visual and textual descriptors. The contributions of this paper are twofold. First, to the best of our knowledge, this work is the first study of the influence of LSA on the retrieval of a significant number of multimedia documents (i.e. collection of 20000 tourist images). Second, it shows how different image representations (region-based and keypoint-based) can be combined by LSA to improve automatic image annotation. The document collections used for these experiments are the Corel photo collection and ImageCLEF 2006 collection. MOTS-CLÉS : Recherche d’information, annotation automatique, sémantique latente, modèle de fusion