Application And Research Of Text Mining Technology In Product Reviews

With the rapid development of information technology and especially the wide use of the Internet,people's life has been greatly changed.More and more consumers can express their views about a certain product online and through their remarks;a certain attitude or sentiment towards the product can be revealed.So,if these comments can be reasonably analyzed and used,businesses and consumers can benefit a lot from them,and finally this will promote the development of our social economy.During the process of analyzing these comments,a lot of texts can be obtained and text-mining becomes the most effective method to deal with these semi-structured or non-structured texts.The ultimate goal of text mining technology in dealing with network product reviews is the classification of text sentiment.And the most important part of the emotional classification is to choose the appropriate classification method.However,the traditional KNN classification method has the problem of large amount of computation and large deviation of the classification.In this paper,a new CPKNN algorithm is designed.Experimental results show that compared with the traditional algorithm,the efficiency and performance of the proposed algorithm are significantly improved.The main research contents are as follows:(1)Analysis Jingdong Apple MacBook Air notebook page DOM tree structure,According to the relevant URL grasping rules,eighty thousand review data can be collected by using Java crawler technology.By pre-processing the text,it can be processed as training corpus.(2)In this paper,the KNN algorithm is improved and an improved CPKNN algorithm is proposed.Firstly,samples with asymmetric density are specially chosen and circularly tailored in order to improve the accuracy rate.Secondly,on the base of the tailored samples,projection pursuit theory is employed to choose smaller and more representative sample database for a lower time complexity of classification algorithm.The values of k are determined by experiments.(3)Combined with CPKNN algorithm,a fine-grained text emotion analysis model is designed,and a product review exploring system was implemented.The system can capture relevant comments on the specified website pages,display segmentation information,extract relevant feature words,classify product comments according to the emotion,and display them in a graphical interface.Results of the system operation have shown that the application of text mining technology in product reviews is effective and feasible.