Supervised self-organization of user's Kansei model for image retrieving

We propose a Kansei-oriented system dedicated to image retrieval. By "Kansei"-oriented, it is understood that the user is at the center of a system sensitive to and actively using his/her cognitive abilities as well as his/her sensitivity. The definition of this sensitivity is very difficult to elicit even for the user him/herself and therefore should emerge from his/her interaction with the system. We propose a dynamic multiagent architecture in which a language, through which the user and the system establish mutual understanding over visual perceptions, develops. We describe the mechanisms underlying the development of this language and which allow the user to have full control, both functionally and structurally, of the system.