Research On Chinese Microblog Public Opinion Analysis

With the constantly innovating of electronic products, the Internethas become an important part of people’s daily life. Microblog relies onits powerful ability of timely communication, a diversity of informationcommunication, large user groups, supports of a variety of terminalequipment, has turned into one of the most popular online social media.Microblog public opinion plays an important role in the Internet publicopinion. Because of the characteristic such as diversity, repeatability,fragmentation of the information on microblog, the users can’t grasp andanalyze the public opinion trends. This article is based on the function ofmicroblog and the structure of the user, study the work of public opinionanalysis. The main research work is as follows:(1) Aiming at the shortcoming of the existing web crawling tools andmethods, a microblog crawling tool based on use’s network structure isdesigned. Starting from user’s network structure, this tool access to themicroblog information resources by simulating user login, constantlyexpand the user queue by fan list of the user and selectively obtain theuser’s microblog comment in the user queue. Useless information isfiltrated by the study of microblog noise information, and the original textdata is pre-process.(2) A method of topic detection and tracking algorithm based on thetopic name is proposed. Microblog can be divided into two categories,known and unknown topic and then the two categories will be processedindividually, the work of new topic detection and topic track is finished;A linear weighing method is designed for calculating the similaritybetween topic and microblog, threshold selection shows the feasibilityand effectivity of this design method; four characters are summarized, amethod of calculating the topic heat is proposed, the hot topic can befound out by sorting the topic heat value.(3) A method of topic name extraction algorithm based on CRFchunk model is proposed. The microblog sentence can be subdivided bybuilding CRF. Modifying the parameter settings by experiment can makethe annotated results best. On the basis of chunk parsing, the name of unknown topic is extracted. Emotional tendency analysis is processing bythe dictionary made up of sentiment word, negation word and degreeadverb. Emotion tendency computing is estimated after analysis.