Research On Algorithms Of Text Sentiment Analysis

With the rapid development of Internet technology in China,more and more fragmentary information is produced,which makes the public opinion monitoring work face new challenges.More and more researchers begin to monitor the fragmentary information through emotional analysis.Aiming at this problem,this thesis mainly uses machine learning method to solve the problem of short text emotion analysis and multi classification and multi granularity emotion analysis under unbalanced data.In the research process of short text sentiment analysis,a CNN-XGB method based on convolutional neural network model and XGBoost is proposed,which solved the problem that the majority of short texts lead to poor results of sentiment analysis model in Internet comments.The main advantage of CNN-XGB method is that it replaces the original classifier of CNN model by XGBoost to improve the accuracy of sentiment analysis of CNN-XGB method.The experimental results show that the proposed CNN-XGB method can significantly improve the accuracy compared with the CNN model when dealing with short text sentiment analysis.The results also show that the CNN-XGB method is feasible.In the research process of solving the problem of multi classification and multi granularity emotion analysis under unbalanced data,a MO-seq2 seq method based on seq2 seq model is proposed to solve the problem of low accuracy of emotion analysis model due to unbalanced training data set in multi classification and multi granularity emotion analysis.The main contribution of MO-seq2 seq method is to optimize the result of emotional analysis directly outputted by the decoding end of the original seq2 seq model into the result outputted by two steps.First,we get the preliminary results of emotional analysis and whether the input data is small sample data.Then we get the results of emotional analysis by combining the results of the previous step.The experimental results show that the accuracy of the MO-seq2 seq method is significantly improved compared with the seq2 seq model in solving the problem of emotional analysis of unbalanced data.At the same time,the experiment proves that the MO-seq2 seq method is feasible.The CNN-XGB method and MO-seq2 seq method proposed in this thesis have some practical significance to solve the problem of emotion analysis.In the future,we can combine the latest model to further study the multi classification and multi granularity emotion problems.