Personalized Tag Recommendation Using Social Contacts

Tag recommendation encourages users to add more tags in bridging the semantic gap between human concept and the features of media object, which provides a feasible solution for contend-based multimedia information retrieval. We study personalized tag recommendation within a popular online photo sharing site Flickr. Contact relationship information of Flickr users is collected to generate an online social network. From the perspective of network topology, we propose node topological potential to characterize its ability of affecting other nodes. With the topological potential metric of the users in contacts network, we can distinguish different social relations between users and find out those who really have influence to the target users. On these social contacts, we acquire the implicit personalized information. Tag recommendations are based on user’s tagging history and the latent personalized preference learned from social contacts. We evaluate our system on large scale real-world data crawled from Flickr. The experimental results demonstrated that our algorithm can significantly outperform the non-personalized global tag co-occurrence method. We also analyze the further usage of our approach for the cold-start problem of tag recommendation.