Public Opinion Monitor System With Text Semantic Sentiment Analysis

With the development of Internet in China,the discussion about social hot topics,cooperation product and political policy has increased much.So the detecting and monitor of online public opinions has been an important area.How to extract topics precisely and how to judge the sentiment in the topics has been the kernel problem in public opinion monitor system.To solve the problems above,this paper analyze the features and requirement based on normal public opinion monitor system.Focused on the shortages of existing public opinion monitor system,this paper design and implement a public opinion monitor system based on text semantic and sentiment analysis,and provides intellectual,professional,and accurate public opinion monitor techniques.To realize the goals,this paper includes work as follows:Analyze the current system from topic acquiring and sentiment analysis aspects based on the detailed environment,and claim the shortages of current systems.Illustrate topic detecting algorithm and sentiment analysis algorithm based on word vectors,as well as their practical usage in the system.Build Neural Network model to calculate the word vectors.Design and implement topic detecting algorithm and sentiment analysis algorithm based on word vectors.Aimed at the current situation that ignores the word context,propose Neural Network model based on word vectors for text sentiment analysis.Aimed at the shortage in description of Chinese syntax features in word vector model,propose and implement sentiment analysis algorithm based on machine learning and word syntax features.Claim the bag-of-words model and the importance of Chinese syntax features.Design the methods to incorporate syntax features into bag-of-words model.Implement 3 machine learning methods: Na?ve Bayes,SVM,and Random Forrest.Finally combine the sentiment analysis algorithm based on word vectors and sentiment analysis algorithm based on machine learning with Boosting tree method,complementing different models.Aimed at the specific requirement for sentiment analysis system,propose architecture design for public opinion system.By dividing the system into online computing component and offline computing component,reduce the online computing cost as well as ensure the real-time user interaction.Design and implement public opinion monitor system,which support the user custom schedule,alert setting,opinion result view,and feedback for opinion results,etc.The result of algorithm and system prove that public opinion monitor system based on semantic and sentiment analysis has a good expansibility,strong profession,less human supervise and accurate result.