Research On Online Public Opinion Recognition Based On Deep Learning

With the rapid development of web technology,the web data grows dramatically.How to efficiently monitor the public opinion in the websites has become an important task.Chinese text sentiment analysis,the key content of online public opinion recognition,plays a crucial role in the web monitoring area.Text sentiment analysis is an important research topic in the area of natural language processing.The traditional text sentiment analysis methods have some drawbacks in dealing with mass data,such as low efficiency and accuracy degradation etc.Recently,the emergency of deep learning technology provides a possible and effective solution to these problems.The dissertation thoroughly investigates the analysis of Chinese text sentiment and deep learning based on the understanding of the related works.In particular,the dissertation proposes a Chinese word segmentation scheme based on the Long Short-Term Memory(LSTM)network model as well as a scheme for the analysis of Chinese text sentiment based on the convolutional neural networks(CNN).The proposed schemes are verified using the typical datasets,respectively.In addition,a system for the practical analysis of Chinese text sentiment is designed and implemented based on the proposed schemes.The main work and contribution of the dissertation are summarized as follows.1.The key technologies in the Chinese text sentiment analysis,i.e.,preprocess,text representation,feature extraction and classification,etc.are introduced.The typical methods for Chinese text sentiment analysis and their cdrawbacks are analyzed.The fact that the key technologies in the Chinese text sentiment analysis are suitable for deep learning is pointed out.The recent research progress of deep learning is overviewed.The typical text feature extraction models are introduced and the neural network language models used in the dissertation are analyzed in detail.2.The performance of traditional Chinese word segmentation methods and programs is degraded when dealing with mass web text data.In order to tackle this problem,a Chinese word segmentation scheme based on the deep neural network model is proposed.The Encoder-Decoder Model(EDM)based on the LSTM network is employed to train the segmentation model,which is used to perform the word segmentation.In order to improve the word segmentation performance,a modification method based on the word vectors is further provided.Experimental results on the typical Weibo dataset suggest that the performance of the proposed scheme is significantly improved compared with the traditional word segmentation software jieba.3.The accuracy of traditional text sentiment analysis methods might be degraded in dealing with mass data.In order to tackle the problem,a method for analyzing the Chinese text sentiment is proposed based on the CNN in deep learning.A method is provided for updating the learning rate in the training process of the CNN in order to improve the convergence performance of the CNN training.In addition,the explanations for the method are provided.Experiment results on the typical datasets indicate that the accuracy of the proposed method can be improved compared with that of the traditional machine learning methods.4.Text features have an effect on the performance and complexity of the CNN-based Chinese text sentiment analysis method.A method for extracting Chinese text features is proposed based on the deep neural network models in the doc2vec program,which can mine the hidden meanings of the texts.Moreover,a normalization method is designed to tackle the nonunifonn distribution problem in the extracted feature values.Experiment results on the typical datasets suggest that the performance of the CNN classification method is satisfactory when the dimensions of the extracted feature are moderate.5.A text sentiment analysis system is designed and implemented based on the proposed scheme.The system design scheme and the modules of the system are described.The implementation results indicate that the designed text sentiment analysis system can fulfill the requirements of practical text sentiment analysis.