No abstract

In the last years, microblogging systems have encountered a large success. After 7 years of existence Twitter claims more than 271 million active accounts leading to 500 million tweets per day. Microblogging systems rely on the all-or-nothing paradigm: a user receives all the posts from an account s/he follows. A consequence for a user is the risk of flooding, i.e., the number of posts received from all the accounts s/he follows implies a time-consuming scan of her/his feed to find relevant updates that match his interests. To avoid user flooding and to significantly diminish the number of posts to be delivered by the system, the authors propose to introduce filtering on top of microblogging systems. Driven by the shape of the data, the authors designed different filtering structures and compared them analytically as well as experimentally on a Twitter dataset which consists of more than 2.1 million users, 15.7 million tweets and 148.5 million publisher-follower relationships.

Social networks have become an important information source. Due to their unprecedented success, these systems have to face an exponentially increasing amount of user generated content. As a consequence, finding relevant users or data matching specific interests is a challenging. We present RecLand, a recommender system that takes advantage of the social graph topology and of the existing contextual information to recommend users. The graphical interface of RecLand shows recommendations that match the topical interests of users and allows to tune the parameters to adapt the recommendations to their needs.

No abstract

National audienceLes services de micro-blogging sont devenus récemment une source d’information importante. Cependant, victimes de leur succès, ils doivent actuellement gérer une quantité sans précédent d’informations générées par les utilisateurs. Il devient par conséquent difficile pour les utilisateurs de trouver dans ces services des contenus proches de leurs intérêts. Afin de recommander des utilisateurs à suivre sur un sujet donné, nous proposons dans cet article des scores basés sur la topologie du graphe social ainsi que sur le contenu textuel des microblogs. Pour permettre le passage à l’échelle, nous présentons une approche qui s’appuie sur l’utilisation de landmarks pour pré-calculer des recommandations pour certains comptes choisis dans le graphe. Nos expériences confirment la pertinence de notre score de recommandation par rapport à des approches existantes ainsi que le passage à l’échelle de notre algorithme basé sur l’utilisation des landmarks