Personalization based on domain ontology
Mehdi Adda, Lei Wu, Yi Feng · 2006
As a consequence of the proliferation of multimedia contents, users are nowadays frustrated with the huge amount of available video information whose content is not targeted to their needs and preferences. Its challenging to analysis video content for video personalization due to the lack of semantic video summarization and retrieval techniques. In fact, most of current video personalization systems are using low-level features. However, users identify and select video content using high-level semantics. This creates a gap between user preferences and video content representation that must be bridged for video personalization systems.In this paper we present a new approach for video personalization based on domain knowledge. We first introduce an ontology based indexation approach to enhance retrieval performance. Then, we present a personalization strategy based on fine grained sequential pattern discovery. The proposed approach is based on both user and content personalization. The performance study and experiments show that the use of ontologies to index and represent video contents enhance running time and memory performances. This paper also describes VideoMiner, a system prototype that implement the proposed approach for video personalization.