Research On User Satisfaction In Music Based On Text Mining

With the rapid development of Internet technology,people no longer only pay attention to material products,but pay attention to all kinds of emotional marketing products.Music,as a carrier of emotional expression and resonance,has attracted more and more attention.However,there are a lot of user comment data in music platform.How to dig out users’ concerns from these comment texts and apply them to the construction of user satisfaction system is of important auxiliary significance for platform to discover different needs of users and timely optimize platform functions.This article selects the Netease cloud music platform and QQ music platform as the research object,through dynamic acquisition program which user comment text,after word segmentation,to stop words after reprocessing operations on the comment text exploratory analysis,mainly from word cloud,semantic network graph model to explore the word frequency characteristic in the comment text,such as mining associations between words.On the basis of the word frequency features obtained,coarse-grained sentiment orientation analysis was conducted from the perspective of sentence hierarchy,and the corresponding sentiment orientation of each comment was determined by machine learning method.In addition,LDA theme model is used to extract theme features from QQ music reviews and Netease cloud music reviews,respectively,to mine the hidden theme categories in the content of music reviews,providing alternative words for text clustering to determine the final category features.In the process of building a system of user satisfaction,first word2vec method to quantify the comment text processing,after the vectorization process after the text of the text characteristic clustering analysis,finally determined by error function and the "law of elbow" the number of categories,according to the result of clustering sums up the indicators of customer satisfaction index system.Statistics on the basis of finishing the index,the frequencies of the corresponding indicators of key size,and calculate the frequency of every index of situation,constructing judgment matrix of each hierarchy index based on the analytic hierarchy process to determine the weight of each index categories corresponding,statistics of all indicators under the different dimensions of the corresponding number of positive,negative and neutral comment text,The proportion of positive comments is taken as the corresponding score of the index,and the weighted sum method is adopted to calculate the final user satisfaction.The research results show that,after the text characteristic clustering features in seven categories: the university entrance exam,one’s deceased father grind,love,family love,friendship,song,emotion,and the ultimate satisfaction score is 0.6375.From the index level score,users for basic attribute class satisfaction score is low.This also indicates to some extent that some users tend to pay attention to the content of music,indicating that there is still a large space for the platform to improve the quality of music content.Based on the results of the satisfaction evaluation system,we analyzed the influencing factors and divided the indicators into four different areas: advantage area,area to be improved,low-priority area and over-supply area,so as to put forward specific improvement suggestions for different index elements.