Research On Sentiment Analysis Method Of Commodity Reviews

With the rapid development of online economy,there are more and more commodity reviews,and the contents of reviews are different.In such a scenario,it is worth analyzing and mining these comments to get the emotional tendency.Because it can not only provide some reference for consumers who need to buy goods,but also help sellers to improve the quality of goods and improve the service levelThis paper focuses on the analysis of emotional attributes based on machine learning and deep learning.Firstly,this paper compares the effects of different word segmentation techniques on the model,so as to find better word segmentation tools.Through the comparison of the word segmentation of jieba,pyhanlp and snowlp,it is found that the effect of the word segmentation of jieba is far better than that of the other two word segmentation techniques,no matter from the speed or quality of word segmentation.Then,based on the segmentation results,Word2vector is used to train the word vector and take the word vector as the input.Then,the machine learning algorithms such as logical regression,SVM,decision tree,and the deep learning models such as LSTM,BI-LSTM,attention mechanism and their fusion are used to build the classification model for the Chinese commodity review data in this paper.Then,we use the precision,recall and accuracy to evaluate the effect of the model.We find out a more suitable emotional classification model for the dataset.In this paper,word vectors are weighted and unweighted,and the results of the model are compared.It is found that unweighted word vectors have a higher degree of discrimination to emotional tendency.This paper also proposes a StackLSTM plus attention mechanism model,which is a neural network of stacklstm plus attention mechanism,to extract higher-level information,and verifies the effectiveness of using this model on a given dataset.