Design And Implementation Of Network Public Opinion Analysis System Based On University

With the rapid development of Internet technology,online life has been an important way of life for contemporary college students.The guidance of Internet public opinion also becomes an important part of college ideological work.The group of college students is in an important period of the forming and developing the view of life,values,and world outlook.They are vulnerable to online rumors,resulting in aggressive behaviors and serious group events.Therefore,it is particularly important to obtain university public opinion data,detect public opinion hot topics,and analyze the emotional disposition of topics.It is especially important for the early detection of public opinion trends,guide public opinion trends in the early stage,and avoidance of major public opinion security incidents.On the basis of the traditional public opinion analysis method,this paper studies the key technologies such as acquisition and preprocessing of public opinion data,text feature representation,topic clustering,ho t topic measurement,and sentiment orientation analysis,and forms a set of more complete functions of the college network.The analysis system realizes and displays hot topic detection,topic sentiment orientation analysis,sensitive word management and statistics,topic trend change,public opinion information retrieval,public opinion information statistics,hot news focus display,system related configuration and user management,and other functional modules.It solves the problems that the current university network public opinion analysis system exists a single source of data,a simple processing method,a low detection accuracy,and imperfect functions.The main research content of this article: through the theme of the web crawler,it obtain the University of Chongqing forum,post bar,Weibo,news network and other sources of public sentiment data.A text feature extraction method based on Word2vec&LDA is studied,which improves the high-dimensional sparseness of traditional text representation based on VSM and ignores the potential semantics.At the same time,the text feature representation combined with Word2 vec and LDA takes into account the LDA text-theme features and word2 vec word space characteristics.In this paper,the Single-Pass&HAC clustering algorithm is proposed based on classic Single-Pass clustering algorithm,which introduces time window and hierarchical clustering to decrease the sensitivity of text input sequence in classic Single-Pass algorithm.Our method makes great improvements bo th in quality and efficiency,and the accuracy of text clustering and topic detection is increased by about 20% compared with the traditional VSM&K-Means algorithm.At the same time,this paper defines a hot topic measurement method in consideration of the actual condition in university,which can be better to calculate the heat value that directly displays interesting topics in students.In terms of text sentiment analysis,this paper expands the sentimental dictionary through Word2 vec on the basis of rules and dictionaries in traditional methods,which solves the problem that basic sentimental dictionary can hardly deal with new words.In consequence,the proposed method greatly improves the accuracy of sentimental analysis.