Summarizing User-Contributed Comments

User-contributed comments are one of the hallmarks of the Social Web, widely adopted across social media sites and mainstream news providers alike. While comments encourage higher-levels of user engagement with online media, their wide success places new burdens on users to process and assimilate the perspectives of a huge number of user-contributed perspectives. Toward overcoming this problem we study in this paper the comment summarization problem: for a set of n usercontributed comments associated with an online resource, select the best top-k comments for summarization. In this paper we propose (i) a clustering-based approach for identifying correlated groups of comments; and (ii) a precedence-based ranking framework for automatically selecting informative user-contributed comments. We find that in combination, these two salient features yield promising results. Concretely, we evaluate the proposed comment summarization algorithm over a collection of YouTube videos and their associated comments, and we find good performance in comparison with traditional document summarization approaches (e.g., LexRank, MEAD).