Gradually adaptive recommendation based on semantic mapping of users′ interest correlations

SUMMARY In this paper, we propose a gradually adaptive recommendation model based on the combination of both users' commonalities and individualities that depend on the semantic mapping of users' interest correlations. We analyze users' information access behaviors and histories to extract users' interests and trace their transitions. In details, according to a set of bookmark tags classified by a semantic means, the pages accessed by users are assigned into several tag classes, which will finally be clustered into different groups in accordance with the types of interests that belong to two categories: personal and common interests, respectively. Based on the detection of users' interest focus transitions through interactions between users, we provide a series of information seeking actions in sequence to the target users. Besides, according to the reference groups which are defined to describe different relations with the target users, the successful experience is extracted and recommended. After the description of the definitions and measures, the mechanism to infer the interest focus, the system architecture and experimental evaluation results are described and demonstrated.