Information extraction based on personalization and contextualization models for multimodal data

Multimedia data concern visual, audio, textual, information involving voice, gestures and others, so there are several challenges related to their storage and retrieval. Context awareness and personalization are key aspects to consider to manage any information-related content. In this paper we propose a general profiling model to define the user and context profiles in order to preserve and use their interrelationships. This general profiling model can be used to filter the user query in a more precise and selective way and to filter the multimedia contents in the repository that match the user query. In this paper the personalization relies on the combined use of the user profiles and context profiles, and the combination driving in particular the choice of the interaction mode in order to make relevant information available to people in the appropriate form.