Research On Intelligent Speaker Online Review Based On Text Mining

With the rapid development of 5G,artificial intelligence,big data and other emerging technologies,technology product with intelligent and IOT continue to enrich and facilitate people's lives,which is choosed by more and more consumers.Intelligent speaker is a representative intelligent product.At the same time,the rapid development of e-commerce platform,as well as the continuous maturity of logistics system,consumers choose to shop online and present online reviews,resulting in a large number of text data.So for the mass of information use natural language processing technology,which is of great significance to excavate consumers' attention to product features,conduct emotional analysis and user satisfaction analysis.Using Python web crawler technology,this paper selects three brands of tmall genie,Xiaodu and Xiaomi as the research objects to mining and analyse the text of the product reviews of the smart speaker with no screen.The main research are as follows: firstly,preprocess the text,draw the word cloud image by brand and extract the theme of the good and bad reviews based on LDA theme model to realize the text feature analysis.Then,use SVM,NB and KNN nearest neighbor machine learning methods to classify the sentiment of good and bad reviews.Finally,on the one hand,based on the results of text feature analysis,determine the product feature words and extend them by word2 vec.And then construct the comprehensive evaluation index system of satisfaction degree and determine weight by AHP.On the other hand,extracte the short sentences,and constructe the emotion dictionary to calculate the emotion value.Based on the time interval between the initial review and the additional review,correct the emotion value by normalization method.And lastly calculate the satisfaction value and compare and analyse them about the three brands.Through the research,the following conclusions are obtained: 1)eight characteristics of logistics,customer service,sound quality,appearance,price,function,intelligence and control reflect the user's concerns,which are used as evaluation indicators;2)consumers give good reviews on appearance and function,but give poor reviews on customer service,price and control;3)SVM classifier based on linear kernel function has the best effect,which can be used for Text classification about good and bad reviews of intelligent speaker;4)in terms of comprehensive satisfaction,Xiaomi speaker has the highest satisfaction which has relatively highest score of multidimensional.The second is Xiaodu,with has colsed score of dimension in appearance,function and intelligence to Xiaomi.And the last is tmall genie,with high cost performance and complete functions.Finally,the following suggesions are put forward: 1)for businesses,it is necessary to stabilize the price,optimize the connection and voice interaction,and strengthen the service awareness of customer service staff;2)for brands: tmall genie should optimize the sound effect,voice interaction,and home control;Xiaodu need to stabilize the connection,improve the sound effect,optimize the Baidu and other music app membership restrictions;Xiaomi also need to stabilize the connection and control the price fluctuation.