An Analysis Of Affective Tendecny Of Short Text Based On Deep Learning

With the rapid development of the mobile Internet,people participate in various network activities through the mobile terminal,resulting in a large number of short texts with emotional orientation.How to quickly find out the emotional orientation of these short texts from the Internet and provide effective help for government,businessman,and individual decisionmaking has become a hot topic in the field of natural language processing.For the key points in the analysis of sentimentality of short texts on the Internet,this thesis mainly studies the following aspects.First of all,aiming at the problems existing in traditional text representation,this thesis used static and non-static ways to train the two public data sets to get the corresponding word vectors and solve the related problems of text representation,based on the unsupervised Word2 vec model.Then according to the problem of text feature extraction,the traditional sentiment analysis method is studied and analyzed.The experimental data are constructed by using word vector and TF-IDF respectively.The experiment is conducted as the input of three classical machine learning models.Comparing the results of the experiments with each other,comes the result that there are limits for the machine learning to learn emotional information contained in word embedding.In order to solve this problem,the deep learning method is introduced into the network short text sentiment analysis.Through contrast analysis,a convolutional neural network model called TCNN,which based on deep learning,is proposed.In order to confirm the effectiveness of TCNN,this thesis compared the experiments between traditional machine learning models and TCNN model.At the same time,Compared and analyzed the effects of factors that may affect the effect of the model.Finally,based on the TCNN model,in order to extract more fully local sensitive information in the text,a two-channel convolutional neural network model called Double-TCNN is proposed.Compared with the traditional machine learning model and the TCNN model,the validity of the model was verified.In summary,on the problem of short text sentiment orientation analysis,this thesis uses the word vector to solve the problem of text data representation.Based on the traditional machine learning model,a convolutional neural network model called TCNN based on deep learning is proposed.Compared with the traditional machine learning models,the TCNN model has better results.To improve the TCNN model,the Double-TCNN convolutional neural network model is proposed.Experiments show that the model is better than the traditional machine learning model and the TCNN model.