Research And Implementation Of Public Opinion Analysis System For Hot Cases

As the development of the information society and the popularization of the mobile Internet,different kinds of information in society are closely related to the Internet.By the aid of the Internet,information can be quickly transmitted to a great quantity of netizen through various channels.As a medium,network can help social public opinion be expressed in cyberspace more quickly and concretely,forming network public opinion.Internet public opinion is a product of the new era.Its expression is not only free and interactive,but also misleading and unexpected.Using Internet public opinion information properly and developing its positive influence are conducive to form a positive and healthy public opinion atmosphere,guarantee social harmony and stability,and improve national happiness index.In the new era,network public opinion has become an important approach to express social emotions and values.Among them,the public opinion of hot cases is closely related to people's daily life,and has aroused wide public concern.The rational use of public opinion information of hot cases contribute to facilitate public supervision,strengthen the control of public opinion,and realize fair and reasonable judicial referee results.Based on this,this article conducts related research on public opinion analysis of hot cases,proposes a series of public opinion analysis models and implements a public opinion analysis platform for hot cases.The main work is as follows:(1)In order to fully express text semantics and improve the quality of natural language processing tasks,Semantic Dependent Embedding is proposed.By fusing semantic dependent information,it can further extract and merge text semantic information on the basis of traditional word vectors.Semantic Dependent Embedding is the basis of the follow-up research in this article,and is widely used in downstream tasks in various scenarios to improve the quality of tasks.(2)In order to fully extract the hidden semantic features of the text and reduce the impact of text length on semantic accuracy,Fusion Feature Extraction Network is proposed to extract the internal details of the public opinion text.The network abstracts text semantics from multiple perspectives and aspects.Besides further improving the accuracy and completeness of semantic expression,it also ensures the accuracy of the polarity analysis of public opinion texts.(3)Aiming at the problem of similar text matching,Deep Semantic Matching Model is proposed based on matching-aggregation architecture.In combination with Semantic Dependent Embedding,Fusion Feature Extraction Network,multiple view Attention and other components,this model matches two sentences from a more fine-grained perspective and judges the similarity,improving the accuracy of the matching results.(4)By using Java SSM framework,Mysql,Tensor Flow framework and other techniques,this article implements a public opinion analysis platform and makes an in-depth analysis of the important characteristics of micro-blog public opinion comments.In the analysis process,the data acquisition module gathers some comment information from the Weibo platform,the sentiment analysis module analyses the polarity of public opinion texts and hot topics,and the similarity matching module retrieves similar legal documents.In addition,the user management module is used to manage account information,authorization information,etc.