Twitter Summarization Based On Sparse Reconstruction And Social Network

With the rapid growth of microblogging services,such as Twitter,a vast of short and noisy messages are produced by millions of users,which makes people difficult to quickly grasp essential information of their interested topics.Twitter summarization aims to generate a succinct summary delivering the core information from a sheer volume of tweets in a given topic.It can help to improve the efficiency of people's access to information,and simultaneously help relevant departments monitor event process and then control the direction of public sentiment.Though document summarization have been researched for many years,it is still a knotty problem due to the large scale noisy,informal nature of messages in social media.Traditional summarization methods only consider text information,which is insufficient in social media situation.Existing Twitter summarization techniques rarely explore relations between tweets explicitly,ignoring that information can spread along the social network.Based on problems mentioned above,we propose a novel approach for Twitter summarization by integrating Social Network and Sparse Reconstruction(SNSR).1)Inspired by social theories that expression consistence and expression contagion are observed in social network,we model the tweet-level network and incorporate it as a social regularization into a sparse reconstruction based framework,which supposes that a good summary can best reconstruct the original corpus.Social regularization can help reduce the reconstruction error and make a rectification during the process of reconstruction,which means that two related tweets can maintain some relevance before and after the reconstruction.2)Two similar tweets will raise the salience of them throughout the corpus in the process of reconstructing each other.It can lead lots of similar tweets present in the final summary.Through incorporating diversity regularization into the sparse reconstruction based framework,we can avoid the ”similar” reconstruction phenomena to remove redundancy.3)The whole framework can be regarded as a mathematical optimization problem,then we develop a Nesterov's Accelerate Gradient(NAG)based algorithm to solve it.4)Finally,we construct the gold standard Twitter summarization corpus,experimental results on this datasets show the effectiveness of our framework.In conclusion,we propose a framework by integrating social network and sparse reconstruction,which considering the text content and structure information.This framework manages to acquire more semantic clues through mining tweet-level potential network.The whole framework solves the salience and diversity problem of summarization research by integrating social,diversity and sparse regularization.It provides an inspiring research idea for Twitter summarization research.