The Research On Micro-blog Sentiment Analysis Based On Deep Learning

With the rapid development of the Internet,social network services have shown explosive growth.An increasing number of people are getting used to expressing their opinions and feelings through micro-blog.It has great application value for sentiment analysis and mining of massive text on micro-blog platforms,and has become a new research hotspot in recent years.The traditional method of text sentiment analysis relies on complicated feature engineering,and it is difficult to adapt to the characteristics of Weibo text in a concise,diverse and constantly changing manner.In recent years,deep learning technology has been more and more widely used in the field of natural language processing.In this paper,deep learning technology is used to optimize the existing micro-blog sentiment analysis methods.Two kinds of deep learning models are designed as follows:Firstly,considering the importance of each word in a microblog sentence to the overall emotional expression of the sentence is different,a BGRU-Attention neural network model is designed by applying Attention Mechanism in deep learning to the BGRU neural network.BGRU can effectively capture long-term relevance features of text.Attention mechanism can give more weight to important words when the model synthesizes high-level emotional features.It is also helpful to improve the interpretability of deep learning model.Experiments show that BGRU-Attention model is more effective than traditional support vector machine-based model and other deep learning model in affective orientation analysis of English micro-blog.Visualization of Attention layer shows that the model chooses words with stronger emotional orientation to give higher weight.In addition,this part also organizes a number of comparative experiments to explore the influence of the quality of pre-training word vector on the effect of the BGRU-Attention model.Then,considering the limited effect of BGRU neural network on fine-grained sentiment classification,a BGRU-CNN neural network model is designed by combining the bidirectional GRU neural network and convolutional neural network.Combining the advantages of BGRU and CNN,we use CNN to enhance the capture and extraction of local important features and enhance the robustness of BGRU.The hierarchical classification method is used to further improve the effect of the model on fine-grained micro-blog sentiment classification tasks.Experiments on NLPCC2014 Chinese micro-blog sentiment analysis data set have achieved better classification results than traditional models and methods.Through a series of comparative experiments,it is shown that deep learning technology can effectively improve the effect of micro-blog sentiment analysis,and the two improved methods designed in this paper further improve the effect of micro-blog sentiment analysis.