| With the growth of social media,sentiment analysis has become a hot topic in natural language processing.The field has been developing rapidly in recent years.Meanwhile,deep learning is bringing along the occurrence of distributed representation method based on artificial neural network.Nowadays,Word2Vec has become the most commonly used model based on distributed representation.Word2Vec model optimizes iterations based on random initial word vectors,so as to obtain stable word vectors.However,if the goal is to solve the sentence vector,it needs to be weighted.Therefore,this paper systematically summarizes the current research on the representation technology of initial word vector and the weighted solution of sentence vector and analyzes their performance and puts forward its own innovative technology.Firstly,the initialization technology of word vectors based on Word2Vec model is improved.On the basis of the original random vector,word frequency information is added to change the distribution characteristics of the initial word vector,improve the convergence speed of the model and enhance the performance of the word vector.In the simulation experiment,the accuracy of emotional polarity classification is 1%to 5%higher than that of the original model.Secondly,it is necessary to apply effective weighting methods to get sentence vectors from word vectors.At present,there are many weighting methods,such as the average value of word vectors,TFIDF and so on.In order to compare the performance differences of each weighting method,this paper also proposes a weighting method based on emotional vocabulary,which is given corresponding weights according to the polarity of emotional vocabulary.Through simulation and verification of various weighting methods,it is found that the weighting method of affective vocabulary has high classification accuracy.Especially in English corpus,the accuracy of classification is improved obviously.Finally,these methods are applied to real-world scenarios to further verify their effectiveness.In addition,through the analysis of Xiamen hotel reviews,the factors affecting customer satisfaction are explored,so as to provide valuable suggestions for hotel operators. |
