Image recommendation on content-based bipartite graph

Abstract

Recently, online social curation networks attract lots of users due to its convenience to retrieve, collect, sort and share multimedia content with each other. And high quality recommendation on social curation networks becomes urgent in current complex information environment. In this paper, we proposed a content-based bipartite graph algorithm for social curation network recommendation. Bipartite graph employs relationships between users and items to infer user-item association for recommendation. Beyond the traditional bipartite graph, we introduce the content of items into bipartite graph to extend the recommendation scope and improve its recommendation diversity simultaneously. Furthermore, content similarity is employed for recommendation reranking to improve visual quality of recommended images. Experimental results demonstrate that the proposed method enhance the recommendation ability of bipartite graph effectively in diversity and visual quality.