Sentiment Analysis Based On Topic Model And Ensemble Learning

As the number of Internet users in China continues to increase and the continuous improvement of Internet penetration rate,the development of major Internet platforms and mobile Internet terminals has gradually matured,making the threshold for people to use the Internet in daily life is constantly lowered.With the massive growth of Internet users,analysis and mining of Internet text data to extract the author's emotional tendency,find the hidden information of users and guide and use it,has a strong market intelligence,public opinion prediction and many other fields.Practical significance.Conducting sentiment analysis on these subjective unstructured texts and extracting the author's sentiment tendencies are of great significance in the fields of public opinion monitoring and market intelligence.Based on the research of some scholars,this paper proposes an ensemble learning classifier based on topic model for sentiment analysis.The main research are as follows:First,because the traditional LDA topic model is based on the bag-of-words model,important semantic information such as word order in the text is ignored.So this paper tries two methods to solve this problem: the first method is to use LDA2 vec topic model for information extraction;the second method is to propose an LDA topic model based on n-gram algorithm.This paper verifies the effectiveness of the two models in the field of sentiment analysis through experiments,which can increase the accuracy of sentiment classification.By comparing the model results,it is concluded that the LDA2 vec topic model has the best effect in the field of emotion analysis.Second,this paper proposes an integrated learning classifier based on LDA2 vec theme model.Based on the sentiment analysis using the LDA2 vec topic model,the sub-sample set is divided based on LDA2 vec topic model for ensemble learning.Considering the advantages of ensemble learning in improving the effectiveness and stability of the model,combined with the characteristics of different topics in the text of the topic model,the sub-training set is sampled based on the topic to provide differences for each base classifier in the ensemble learning.The simple voting method is used for decision fusion to obtain the final sentiment analysis result.The analysis of the experimental results shows that using the ensemble learning classifier based on the LDA2 vec topic model proposed in this paper for sentiment analysis can effectively improve the effect and stability of sentiment analysis.