Research On Sentiment Analysis Methods Based On Transfer Learning

The development of social media has a profound influence on the way people live.At the same time,network users have left a large amount of text data on these platforms.It is increasingly important to mine useful information.Text sentiment analysis,as a method of textual sentiment orientation and opinion mining,has gained more and more attention.The existing text sentiment analysis methods can be mainly divided into emotional dictionary based method and machine learning based method.The method based on sentiment dictionary neglect the context and the order of words in the text.And there is no large amount of high-quality tagging data to train a high-precision machine learning model for sentiment analysis.In order to solve these problems,the method of sentiment analysis based on transfer learning was proposed.There are still many deficiencies in knowledge transfer of the methods based on machine learning in different sample distributions,and the negative transfer cannot be avoided well.This paper proposes a sentiment analysis method based on improved Adaboost and transfer learning with Gaussian Processes for these problems.In this method,the most similar part of the source domain data to the target domain data distribution is selected.Secondly,the selected data is used as the final training sample to establish a Gaussian Processes classification model,and the knowledge transfer is realized in this process,Finally,the test data of the target domain is input into the classification model for sentiment classification and output classification labels.The experimental results show that the sentiment analysis method based on transfer learning is more effective than the traditional machine learning method,and the proposed method has a higher accuracy than the one without prior screening of source domain data.Moreover,this paper proposes a sentiment analysis method based on Kmeans and online transfer learning.Firstly,the source domain data and a small number of target domain data are clustered,and the source domain data which is similar to the target domain data is selected to establish the source domain classifier.And when the target domain data is come,a new classifier is established,and the source domain classifier and target domain classifier are combined with appropriate weights to realize knowledge transfer through online transfer learning algorithm and complete the final emotional classification task.This method can not only avoid the problem of negative transfer in online transfer learning,but also transfer offline knowledge into dynamically updated data and perform sentiment analysis on the dynamic data.