Quality biased multimedia data retrieval in microblogs

With the rapid development of social media platforms, huge amount of user generated contents (UGC) are generated ceaselessly. In recent years, content based microblog retrieval has attracted extensive research attention. Effective microblog retrieval services complex analysis of short text and multimedia contents. In this paper, we present a quality biased multimedia microblog retrieval framework. First, we develop an anchor graph based multiview embedding framework which maps the multimedia content features into a unified latent space. Then, the content matching scores of testing microblogs related to the query are obtained by a Markov random field. Further, we employ an quality model to incorporate both microblog quality and content matching. As compared with the state-of-art methods, experimental results demonstrate the effectiveness of the proposed approach.