Research On Chinese Weibo Negative Sentiment Analysis Method Based On Attention Mechanism

With the rapid development of Internet technology and social networks,the number of weibo users is also increasing.In the era of big data,mining the emotional tendency of network text through artificial intelligence technology not only helps to timely understand the network public opinion,but also plays a great role in improving the experience of microblog products and improving the attention to users' personal emotional status.As a part of artificial intelligence technology,emotion analysis technology is of great significance to obtain the emotion trend of users' blog posts.In most of the current studies,the emotional tendency of the text is often divided into two categories(positive emotion and negative emotion)or three categories(positive emotion,neutral emotion and negative emotion),with a lack of attention to the negative emotion of the user.For the task of categorizing negative emotions in text,this paper mainly does the following work:(1)This paper adopts a bidirectional short and long term memory network model based on attention mechanism.Network model in order to validate the bidirectional both short-term and long-term memory in the negative emotion classification task more capture the effectiveness of the context information,under the same conditions,recurrent neural network(RNN),convolution neural network(CNN),both long short-term memory(LSTM),gated gecurrent gnit(GRU),bi-directional gated recurrent unit(BiGRU)and bi-directional long short-term memory(BiLSTM)compares the classification effect of negative emotions,the experimental results show that BiLSTM model has higher accuracy,and different negative emotions of F1 values are higher than other models,It shows that the model can effectively extract the upper and lower information and make effective prediction for different negative emotions.(2)since the research in this paper focuses on negative emotions in the text,and in order to further improve the accuracy of the classification of negative emotions,the attention mechanism is analyzed and studied in this paper.Through the comparison experiment of the fusion of attention mechanism and BiLSTM,it was found that the accuracy of the BiLSTM model was greatly improved compared with the BiLSTM model without the fusion of attention mechanism.Among them,the BiLSTM model based on the self-attention mechanism had the highest accuracy,reaching 83.5%,and was more accurate in the screening of different negative emotions.It shows that the performance of deep learning network can be significantly improved by combining attention mechanism in the analysis task of negative emotion.(3)In order to explore the performance of the best bidirectional long and short term memory network model based on self-attention mechanism in the above work on different data sets,this paper set up an open test experiment,and tested with additional data sets without adjusting the model parameters.The experimental results show that the accuracy of classification is still at a high level,indicating that the model has good applicability and robustness.