Users Are Known by the Company They Keep: Topic Models for Viewpoint Discovery in Social Networks

摘要: Social media platforms such as weblogs and social networking sites provide Internet users with an unprecedented means to express their opinions debate on a wide range of issues. Concurrently growing importance in public communication, may foster echo chambers filter bubbles: homophily content personalization lead be increasingly exposed conforming opinions. There is therefore need for unbiased systems able identify access varied viewpoints. To address this task, we propose paper novel unsupervised topic model, the Network Viewpoint Discovery Model (SNVDM). Given specific issue (e.g., U.S. policy) well text interactions from discussing site, SNVDM jointly identifies issue's topics, users' viewpoints, discourse pertaining different topics In order overcome potential sparsity network (i.e., some interact only few other users), extension based Generalized Polya Urn sampling scheme (SNVDM-GPU) leverage "acquaintances acquaintances" relationships. We benchmark proposed models against three baselines, namely TAM, SN-LDA, VODUM, viewpoint clustering task using two real-world datasets. thereby evidence that our model its SNVDM-GPU significantly outperform state-of-the-art show utilizing greatly improves performance.