Analysis And Research Of Network Public Opinion Based On Big Data

The rapid development of the Internet gives individuals the freedom to fully express their own opinions and value appeals,and also accelerates the emergence and evolution of online public opinion.People are more and more accustomed to express their views on hot events on online platforms,which makes online public opinion gradually become a new social force and public opinion space.The rise and fall of network emergencies has made all circles of society pay more and more attention to the discovery,guidance and governance of online public opinions.How to timely observe the changes of online public opinion in the era of big data,so as to prevent the public opinion from evolving into public opinion and solve the potential public relations crisis has become an increasingly concerned issue for managers.At the same time,the methods of natural language processing are gradually becoming mature,which makes the development and updating speed of sentiment analysis technology for text become unprecedented.With the improvement of computational power of GPU used for accelerated parallel training,deep learning has once again become a hot technology that people pay attention to.Cyclic neural network,convolutional neural network and various improved models have been widely used in various natural language processing tasks.In this dissertation,natural language processing technology is applied to the research of network public opinion analysis.The content mainly includes the following three aspects:(1)With the microblog text as the main collection targets,a big data collection tool was built,and the text collection task was managed and scheduled by using Scrapy crawler framework,Webdriver was used to simulate access to sina weibo webpage,and XPath was used to extract the blog content from the webpage.Text was saved to the Mongo DB database.Batch collection of microblog public opinion text was realized.(2)Based on Tensor Flow,LSTM,GRU,DPCNN and other deep learning models were established,and the performance of each model was compared and analyzed through experiments.On this basis,an improved model BGDA based on model fusion method was proposed.the BERT pre-training model was used to embed words of text as the Embedding layer.BiGRU and Attention mechanism were used to solve the long-term dependency problem,DPCNN was used to extract the context information of the text,and the emotional tendency of the text was determined by combining the output of the two parts.The model is verified by experiments.(3)The constructed big data acquisition tool was used to collect the text of hot topics on microblog,and the crawled data was cleaned and word segmentation was processed.The BGDA model was used to analyze the sentiment tendency of online public opinion texts.