| With the development of the Internet age,Weibo has become an effective channel for people to get information and communicate emotionally.Users publish and spread information on Weibo,which makes Weibo accumulate a huge amount of information data.These data show different people’s views on different events,have obvious emotional tendencies and personal positions,and form an obvious public opinion tendency in cyberspace,which has very rich potential value.In order to reduce the extracted feature dimension and enhance the understanding of context semantics,this paper uses the TF-IDF weighted Word2 vec word vector generation language model as input,and compares the performance of the fusion model and the single deep learning model with the Weibo comment data crawled by the octopus collector.In order to improve the efficiency of model training and the degree of semantic understanding,three different deep learning models are integrated on the basis of LSTM model and the optimal model is selected,which can provide early warning and supervision for the occurrence of hot events and social sentiment,and provide reference for enterprises and governments to understand the market and social sentiment.The research method used in this paper is based on long-term and short-term memory network(LSTM)and combined with convolutional neural network(CNN)to improve the feature extraction ability of the model.Bi-LSTM is fused to capture longer-distance dependencies.In the training of language model,the word vector method combining Word2 vec word vector model and TF-IDF algorithm is adopted.The experimental results show that the LSTM-based feature fusion model has a better classification effect than other models in the emotional classification of Weibo comments. |
