Research And Implementation Of Fine-Grained Topic Evolutionary Analysis Based On The Probability Model

In recent years,the rapid development of online social platforms such as Weibo,Twitter,Facebook and the like has surged the number of users and hit the traditional news media,becoming the most important and convenient source of information and communication channels for people in today's society.People also study and rely on theories and methods of online social network analysis to discover and analyze topics implicit in social media and their evolution.With hundreds of millions of people from different backgrounds and with different backgrounds,they continuously share their opinions and thoughts on the thousands of social hot issues in social media.The demand for evolution analysis of the above topics is also forced to be precise And more granular development.The evolution of fine-grained topics based on the content of the user's interest in discovering and tracking events appearing in social networks has become one of the important issues in current topic evolutionary analysis.Based on the research of topic discovery and evolution,this dissertation studies deeply on the fine-grained topic evolution model based on the online social network text stream.The main focus is on how to mine high-quality,fine-grained topics of interest to users from highly dynamic online short text data streams and to track and study their evolutionary processes.Help users effectively understand and grasp their concerns in a rapidly changing network era the most cutting-edge,most sensitive topics,and according to the evolution of these topics and trends to make the appropriate judgments and decisions.The main contents of this paper include the following three parts:First,the current mainstream topics of discovery and evolution of technology research,combined with Weibo and other online social networks in the short text and the characteristics of the document collection comes with time information,mostly for the user more focused on topic evolution analysis,more details This paper analyzes the existing problems in the solution of this problem and its limitations.Based on the TTM(targeted topic model),this model is extended to a fine-grained topic evolution model that can analyze the topic evolution of online text streams FG-TEM(fine-grained topic evolution model).Second,it defines five kinds of relations between subtopics.By dividing the document set into fixed-size time windows,the document-topic probability distribution and topic-word probability distribution in different time windows are obtained through the fine-grained topic evolution model.By calculating the KL divergence between subtopics in different time windows to determine the similarity between subtopics,the evolution paths of subtopics are obtained according to the relation of subtopics,and then the topic strength of each subtopics within different time windows are calculated.Based on the evolution path of this series of subtopics and the strength change process of subtopics,the fine-grained topic content and strength evolution diagram of the finegrained topics of interest to users in the entire time domain are drawn.Thirdly,according to the characteristics of online social network texts,a new metric of topic coherence E-PMI is proposed.Based on the E-PMI,the proposed fine-grained topic evolution model is evaluated.The experimental results demonstrate that the topic generated by the fine-grained topic evolution model has high coherence,good quality and close to the user's needs.Therefore,the effect of the model better.