Design And Implementation Of Public Opinion Monitoring System Based On LSTM And Attention

With the development and progress of society,the diversification of media and the rapid development of the Internet,the Internet,as a new communication medium and information dissemination medium,has become the main carrier of public opinion information.Whether it is a major domestic or foreign event or a hot spot or focus issue in real life,network public opinion can be formed immediately,and many netizens participate in it,thereby generating network public opinion effects and continuous fermentation.This shows that it is of great significance to help government agencies and companies understand and analyze relevant public opinion in a timely manner.This article first discusses the source,significance and characteristics of online public opinion,and analyzes the characteristics and deficiencies of existing public opinion monitoring software at home and abroad.Based on this,this article introduces advanced technologies such as natural language processing and neural networks to implement a network public opinion monitoring system,which mainly includes a public opinion overview,hotspot analysis,sensitive content analysis,thematic public opinion,public opinion warning,emotional tendency analysis,and real-time monitoring Function module,the front end uses C# language development,and the back end uses Python language development.The system mainly uses focused reptiles to achieve massive information collection on specific topics;hotspot extraction by calculating heat values;keyword extraction using TextRank algorithm more suitable for public opinion system and visualization through word cloud using WordCloud library;three-category emotions are realized Tendency analysis,and compared with LSTM and LSTM model with Attention mechanism introduced,the LSTM_Attention model with the best classification effect is selected.Finally,the three-class sentiment analysis based on LSTM_Attention was realized through jieba word segmentation,Word2 Vec constructing word vectors,etc.,and visualized through pie charts and line charts,so as to obtain an overview of sentiment tendency,the trend of sentiment tendency and the trend of public opinion trend.Finally,the core modules of the system are tested one by one to check the integrity and effectiveness of the system.Practical testing shows that the system can automatically collect and analyze public opinion information,which greatly saves manpower and material resources,can quickly view hotspot issues and heat values,solves the narrow problem of binary classification in sentiment analysis,and greatly improves the accuracy of sentiment analysis Sex,etc.,and can assist government agencies,companies,and companies to keep abreast of public opinion trends and analysis results in order to make corresponding decisions as soon as possible.