Research And Application Of Sentiment Analysis Model Based On Weibo Text

As a social platform that allows users to share and obtain information,Weibo has attracted a large number of users.Weibo text come from all aspects of society.The sentiment analysis of Weibo text will help relevant government departments to grasp the current social public opinion trends in a timely manner,and actively guide and control the development of public opinion based on the results of sentiment analysis,which is important for maintaining social harmony and stability.It is of great significance.Weibo text are short and concise,highly colloquial,irregular in grammar,and serious loss of semantic features.In addition,there are more and more new words that people express emotions,and traditional sentiment analysis models need to be improved.Starting from improving the effect of Chinese Weibo sentiment analysis,the traditional sentiment analysis model is improved,and a sentiment analysis model based on Weibo text is proposed.This thesis mainly conducts sentiment analysis on Weibo text.The main research work includes the following parts:(1)Provide a method to introduce the Attention mechanism on the BiLSTM model.This method uses the word vector model to convert the Weibo comment text into lowdimensional word vectors,uses the BiLSTM model to extract the semantic features of the text,and introduces the Attention mechanism to calculate the corresponding attention features of the text,so that the model can pay more attention to important features Finally,the softmax function is used to calculate the output vector of the Attention layer to obtain the sentimentality of the Weibo comment.Experiments show that the method is effective.(2)A method of introducing the ALBERT model on the Attention-BiLSTM model to extract the semantic features of the text is proposed.The ALBERT model introduced in this method is used to train the word vector as the word vector embedding layer to further optimize the word vector of the input text,and then use the BiLSTM model as the main body of the network to fully learn the text features,and finally use the Attention mechanism to highlight the sentiment analysis of different word pairs.The sentiment analysis of the Weibo comment text is then carried out.The experimental results show that the ALBERT-Attention-BiLSTM model proposed in this thesis can extract the semantic features of the microblog text very well,and the accuracy of the model using the traditional Word2 Vec model and the BERT model as the word embedding layer in the comparative test is improved.In addition,different activation functions are used to test the effect of the model.The experiment shows that the sigmoid function is more suitable as the activation function of the Weibo sentiment analysis model.(3)With the model proposed in the thesis as the core,a sentiment analysis system for hot topics on Weibo was designed and implemented.First,the feasibility and functional requirements of the system are analyzed.Then carry out the overall design of the system according to the demand analysis of the system.Finally,the detailed design and realization of each functional module of the system are described in detail.