The Research Of Text Sentiment Analysisbased On The Fusion Of Dictionary And Word Vector

In recent years,with the rapid development of Internet technology,the number of various social platforms and e-commerce platforms has increased dramatically.It has become a frequent behavior in people's lives to express opinions and emotions on the Internet.Using natural language technology to analyze the various texts published by users on the Internet and to dig out the emotional tendency contained in them is an important way to grasp the social status quo and event dynamics and to obtain after-sales information.Therefore,studying text sentiment analysis methods has important commercial value and social significance.There are still many shortcomings in the traditional methods of sentiment analysis.The existing sentiment lexicons have low completeness,and each field has different emotional words.The general sentiment dictionary is difficult to achieve the desired effect.The traditional word vector technology is obtained according to the context,contains semantic grammar information,but lacks emotional information,and can not solve the emotional analysis task well.In view of the above problems,this thesis uses the constructed domain sentiment dictionary to integrate emotional information into word vector and combines the convolution neural network idea to study the sentiment analysis of online comment text.The main work involved in this article has the following aspects:(1)Aiming at the problems of the completeness of existing sentiment dictionaries and the poor adaptability in the field,the existing sentiment dictionaries are integrated,and using the improved SO-PMI algorithm and mobile phone comment corpus to construct a domain sentiment dictionary in the field of mobile phone commentary.(2)The learning of traditional word vectors is a product obtained by training the learning language model.The distributed expression of words contains only semantic and grammatical information,lacks the problem of emotional information.The constructed domain sentiment dictionary is used to generate the sentiment word vector and merged with the traditional word vector to form an extended word vector containing both semantic information and emotional information.(3)Taking the extended word vector as input,on the basis of the convolution neural network sentiment analysis model.Aiming at the problem that the full connection layer in convolution neural network can not effectively classify the nonlinear distribution data,this thesis presents a comment text emotion classification model based on the combination of convolution neural network and support vector machine.Finally,the experimental results show that the word vector technology of fusion emotion information and the sentiment classification model of CNN-SVM comment text in this thesis have improved the sentiment classification performance in the field of mobile phone commentary.