| In recent years,online Chinese communities,such as Baidu Post Bar,Tianya,etc,provide a platform for millions of online users with common hobbies,where they can share and disseminate instant and meaningful messages about events that occur around them.These messages contain valuable and useful information on a extensive range of topics,including technology,entertainment,stars,movies and so on.However,the massive volume and high variety of data prevent users from sorting out the development of events.The dynamic nature of events also makes it difficult for users to capture and understand the progress of the topic.It is essential to facilitate users to understand the development of incident quickly and easily.In this paper,we study the problem of generating real-time story line to enhance user experience.The main contributions of this paper are as follows:Firstly,we apply a combination of NLP techniques to filter redundant data,which can reduce the difficulty of constructing graph model.In addition,we improve the accuracy of text similarity calculation by considering factors of aging,socialization and semantics.We use the AS-BS based retrieval model to calculate the relevance of topics,thus give more precise summary.Secondly,we model the multi document summarization problem as convex quadratic optimization problem.To deal with streaming data,we propose an online simplex method to generate a real-time text summarization.Finally,based on the generated text summarization,we present to visualize the storyline by a directed Steiner tree algorithm.An online directed Steiner tree algorithm is presented to ensure that the story line is constantly updated.The rationality and effectiveness of our algorithm is verified by comprehensive experiments and user evaluations on real data sets. |
