Application Of Text Sentiment Analysis Based On Attention Network In Public Opinion System

With the popularization of social network and e-commerce,more and more evaluative texts are brought about.Different users use them to express their personal emotions,and how to obtain valuable information from the flooded texts has become one of the focuses of researchers.The task of text sentiment analysis is to judge the emotional color including the evaluative text,based on this,makes this task has received the widespread a ttention in recent years,a large number of scholars adopt different analysis method in order to obtain higher judgment accuracy,some enterprises also applies text sentiment analysis tasks to public opinion analysis system as one of the digital strategy.Therefore,it is of great application value to solve the task of text sentiment analysis and put it into application.In recent years,the application of deep learning technology in image,text and other fields has developed rapidly.Using deep learning to solve the task of text sentiment analysis has attracted extensive attention of relevant researchers.However,the effects of many previous methods are not good,and the text that needs to be paid attention to in different contexts and the interaction between contexts are not well implemented,which limits the effectiveness of different solutions to some extent.In this paper,we propose an efficient new method,mainly using a variety of attention network,and improve on this basis,proposed the iterative multi-attention network,its performance is better than the existing baseline method.The trained method is applied to the public opinion system to further verify the practicability of the method.The main work and innovations of this paper are as follows:(1)In view of the current processing text level is not fine enough,this paper from the perspective of fine-grained analysis,by analyzing the sentiment of aspect words in the text,in order to more accurately judge the text's sentiment tendency.(2)To solve the problem that the semantic characteristics of word embedding used in the current text sentiment analysis task are not rich enough,we take the word embedding of BERT in the pre-trained language model as the text representation of the task.At present,BERT is rarely used in natural language processing.As an emerging technology,relevant experiments have proved its effectiveness.(3)To solve the problem of low performance of traditional sentiment analysis methods,the traditional methods are abandoned and multiple attention networks are used to complete tasks,including conventional attention mechanism,multi-attention mechanism and multi-self-attention mechanism.The final results of the experiment showed that it was better to combine multiple attentional mechanisms.(4)The application of the method of sentiment analysis based on attention network in the actual project of public opinion analysis system further proves the practicability of the method.The realization process of the public opinion system is as follows: Firstly,the crawler technology is used to collect the required data in the corresponding website.Secondly,the data is pre-processed and word segmentation.Then the improved sentiment analysis method is used to judge the polarity of text emotion.Finally,the results are presented visually.Public opinion system can not only help decision makers to analyze problems,but also help users to understand the actual application of different products.Through a large number of relevant experiments,the accuracy of the method proposed in this paper is close to 90%,which can better reflect the tendency of the analyzed text.In addition,the method is applied to the actual public opinion project,and the experimental results pass the black box test and are better than the previous analysis method,which shows that the method proposed in this paper is effective and practical.