Research On Micro-blog Sentiment Classification Based On Hybrid Neural Network

With the continuous development of modern information technology,the world has stepped into the era of the "Internet+" and "web 2.0".The number of users using social media is growing rapidly,and the amount of information and data transmitted on various social platforms is also growing exponentially.As a fast-growing social network platform in recent years,micro-blog provides a convenient cooperation mode for mobile applications,and meets the needs of users in diversified mobile terminals to share information anytime and anywhere,so it is widely praised and used by many users.The analysis of micro-blog text through sentiment classification technology is of great help to understand the public opinion trend,understand the needs of users and improve economic benefitsIn order to improve the classification accuracy of Chinese micro-blog and solve the polysemous phenomenon in Chinese micro-blog,this paper proposes a method of Chinese micro-blog sentiment classification based on hybrid neural network.In the aspect of feature extraction,a variety of representative features are fused.In terms of model construction,a neural network model combining convolutional neural network and recurrent neural network is proposed,which can comprehensively analyze local and sequential information in the text data.The research work and innovation points of this paper are as follows(1)In view of the inadequacy of text feature construction,this paper proposes a feature construction method of mining text information from multiple perspectives.Firstly,the word embedding feature is pre-trained by BERT model on the corpus.Because of the pre training method,this feature can represent the polysemy of the same word in different contexts.Then extract other features,including n-gram and word polarity score features.Finally these features are fused.In order to prove its reliability,this paper designs a comparative experiment of each feature combination.The experimental results show that the feature fusion method adopted in this paper can effectively represent the specific meaning and deep semantic information of the text.(2)Aiming at the problem that the accuracy of sentiment classification in Chinese micro-blog is not high enough,a hybrid neural network model combining convolutional neural network and Recurrent Neural Network is proposed.The model trains both convolutional neural network and recurrent neural network,and then combines the output of convolutional neural network with the output of the last hidden layer of recurrent neural network,so that the output of hybrid neural network can obtain the local representation information of the words,and at the same time,keep the sequential relation between the words.In order to prove the effectiveness of the proposed innovation,this paper designs six groups of comparative experiments,using S VM,CNN,MLCNN,LSTM,c-lstm and hybrid neural network models.The experimental results show that the hybrid neural network model proposed in this paper has higher classification indexes than other comparative models in dataset.