Research On Automatic Summarization Algorithm Based On Topic Detection Method In Social Network

Due to the rapid development of society and the constant destruction of nature by people,natural disasters have occurred frequently in recent years.Therefore,the detection and the rapid support of natural disasters are particularly important.Furthermore,the information interaction is gradually developing into new types of online social media with the developed internet technology.Sina Micro-blogging is such a platform that can make people freely express speech and report news or events in real time which is favored by most people.We can grasp the development of the disasters and the direction of public opinion by analyzing the events or comments on natural disasters published on Sina Micro-blogging.However,the number of information every day generated on Sina Micro-blogging is billions.The content of the blog is very short and the noise data is countless.Therefore,it is very important to detect and extract such specific events.Due to the traditional text analysis technology is not fully applicable to such problems,the microblog-based topic detection and automatic summarization technology has become a hot topic in recent years.(1)Aiming at the problems such as the accuracy of the traditional algorithm is low and the topic representation result is not intuitive,we propose a novel Topic Detection method based on Graph Analysis(TDGA)which utilizes community detection method to detect topics in feature words graph processed by Micro-blog data.In addition,considering the particularity of Sina Micro-blogging,we propose a feature words filtering model and graph generation algorithm to meet the dual requirements of topic detection and community detection.We validated our approach on the natural disaster dataset collected on Sina which has about three thousands of posts.The experimental results clearly reveal the relationship between feature words and natural disaster topics and confirm the scalability and accuracy of the method.(2)Aiming at solving the problem of scattered topic content and semantic inconsistency,a novel Topic-based Automatic Summarization Algorithm(TASA)is proposed to display the final result of topic representation in one sentence.First,we use the sentence sorting algorithm based on topic words and feature words obtained form topic detection method to filter and sort the original Micro-blog data which is the most important step.Then,we selectthe sentence with the highest score and utilize the retouching process to improve the conciseness and richness of the sentence.Therefore,we can use such sentences to implement topic-based automatic summarization while describing each topic.The experimental results show that the topic-based automatic summarization algorithm reflects the exact relationship between topic sentences and the natural disasters which contains rich semantic information.More importantly,we can grasp the basic elements of these natural disasters from topic sentences to help the government guide the disaster relief.