Continuous Summarization for Microblog Streams Based on Clustering

With rapid growth of information found on microblog services, dynamic summarization of evolving information has become an important task. However, the existing work on continuous microblog stream summarization cannot effectively work due to the enormous noises and redundancies. We tackle this problem using a two-step process, first by clustering online microblog streams and maintaining cluster feature vectors. Then, the dynamic summaries of arbitrary time durations are generated from the microblog cluster features. This helps users to better find the worthy interpretations of the online microblog streams. We make use of features to calculate the importance of similar sentences in each cluster for these two steps. Our approach integrates these cluster information with an unsupervised topic evolvement detection model, and illustrate that latent topics to capture the feature dependencies summaries with better performance. Finally, the experimental results on real microblogs demonstrate that our summarization framework can significantly improve the performance and make it comparable to the state-of-the-art summarization approaches.