Flickr group recommendation via heterogeneous information networks

Yueyang Wang,Yuanfang Xia,Siliang Tang,Fei Wu,Yueting Zhuang
DOI: https://doi.org/10.1145/2808492.2808512
2015-01-01
Abstract:As an important characteristic of social media (i.e. Flickr or Facebook), user communities or groups are beginning to attract increasing attention. Most of the previous studies on group recommendation only consider a homogeneous relationship, such as the rating of users to groups in social network. In real world, each group has its textual descriptions (e.g. the \"Sky & Cloud\" group in Flickr), and the describing words about a given group may mention particular entities as well as their belonging categories (e.g. \"Nature\"). In fact, the group recommendation can be conducted in a heterogeneous information network, where informative cues are in general multi-typed. Motivated by the assumption that different types of relationships in heterogeneous information network can be potentially used to boost the performance of the group recommendation. We propose to combine heterogeneous relationship information for users with implicit user feedback. In general, the group recommendation in this paper is formulated as a non-negative matrix factorization (NMF) method regularized with user-user similarity via heterogeneous information networks. Experiments show that our proposed approach outperforms other counterpart recommendation techniques.
What problem does this paper attempt to address?