A Research Of Hot Topic Evolution Analysis

As the Internet is integrated into our lives,social media has become a source of information for us.Based on social media,analyzing the evolution of hot topics will help us quickly grasp the focus of public opinion and clarify the context and trends of topic.It has important practical significance.This thesis studies the evolution of hot topics based on Weibo and describes the evolution process of topics from two aspects: topic content and topic intensity.The main contents include:(1)A topic evolution analysis framework based on event semantics.The framework generates a topic content evolution graph with events as nodes through the event detection method based on neural network and semantic similarity analysis.Combined with topic intensity evolution,our framework describes the evolution process of topics.(2)Event detection and summary method based on neural network.The current event detection method fails to make effective use of text semantic information.We propose a neural network model based on the encoder-decoder framework.The model uses hashtag as the auxiliary information of the event.By using the memory network to generate the vector representation of the event,the model generates the event probability matrix by calculating the similarity between the event vector and the weibo vector.We then detect events by the event probability matrix.According to the analysis of experimental results,the performance of the model has been significantly improved.(3)Topic evolution analysis based on content and intensity.Aiming at the problem of inaccurate excavation of event association relationship in traditional methods,we propose an event evolution relationship recognition method based on word vector and attention.Combining the factors of microblog intensity,the calculation methods of microblog and topic strength are proposed,which effectively analyzes and describes the evolution process of topics.Experiments show that our method improves the performance of event evolution relationship recognition,and the topic intensity change curve is consistent with the current mainstream research.In summary,this thesis carries out a variety of researches on topic evolution analysis.We can finally get the evolution process of topic content and intensity.It has great significance to the areas of public opinion management and information analysis.