Text Sentiment Analysis Based On Deep Learning

In 1940 s,scientists invented the neural network model to imitate animal nervous system.Now with the rapid promotion of computer performance,and the break of traditional problems about neural network,Neural network has been reborned as an old topic.Depth neural networks heap up the activation functions,and the parameters of each activation function are trained by the model,and some complex functions are simulated with the entire network structure.The depth neural network has the advantages of high precision and good generalization performance,now it has made some achievements in all fields.Including image processing,pattern recognition,search engine and other systems.The use of deep learning technology to study natural language processing is both a research hotspot and has a lot of difficulties.The subject of Natural Language Processing,which belongs to the interdisciplinary of computer and linguistics,aims to study the use of computer technology to deal with the text data produced by human beings.The ultimate goal is that people can use various natural languages to communicate with computers without obstacle.The main research directions include semantic analysis,Machine Translation and so on.Sentiment analysis is an important topic in the field of natural language processing foundation,which mainly solves the problem of natural language understanding of the text,including appraise understanding,subjective and objective tendency of understanding,has important significance in the public opinion monitoring,man-machine dialogue etc..Because of the complex form of natural language,it is easy to produce ambiguity,and it is constantly developing and changing,which brings many difficulties to solving this kind of problem by programming.On the other hand,because of the particularity of Chinese itself,compared with the world's main text sentiment analysis,the sentiment analysis of Chinese text has some particularity,there is text vector establishment and Chinese segmentation and many other difficulties,there are lots of problem to study about Chinese text sentiment.This method is based on deep learning,combined with the Chinese text features in-depth study of text sentiment analysis,mainly in the aspects of the work:1.About natural language processing especially the sentiment analysis has been done indepth research and introduce,including the history and present situation of the research,the commonly used method of text modeling,as well as the emotional analysis method based on statistics and the traditional machine learning model,and its corresponding algorithm flow,advantages and disadvantages.2.In view of the special of the Chinese Natural Language Processing problem,this paper thinks that the Chinese text should not divided into word by a certain algorithm,and the character vector is used to model the text directly.At the same time,a theoretical conception for the problem is put forward,and corresponding experimental verification is carried out.3.Proposing a quality evaluation method for Chinese word feature vector modeling,putting forward to evaluate the quality of Chinese word vector quantificationally.4.The TRE-CNN depth learning model structure for Chinese text emotion analysis is proposed.The model is composed of three modules,which are text embedding,feature convolution and text classification.Each module has a certain independence.It has certain reference value for the Chinese Natural Language Processing task.5.A series of related experiments were carried out to adjust the model's super parametric.Comparing with the text sentiment analysis model in recent articles.Experiments show that TRE-CNN model has good classification effect and has wide applicability.