Research On The Utility Evaluation By Theme Of Online Reviews Of Experiential Commodities

With the development of the Internet,online shopping as a new way of shopping gradually accepted and evolved into a way of life.However,on one hand online shopping bring people convenience,on the other hand a serious information gap have been forming because of the Internet space gap,which became a harassment during the process of online shopping.Online comments having been an effective means for buyers and sellers to reduce information asymmetry.Not only because Online comments in the form of graphic combination words,but also because it is from the perspective of the consumers who has purchased the goods to describe the goods.However,due to the accumulation of time,the number of online comments has increased dramatically.The uneven quality of the comments has generated the problem of information overload,so that the reviewers can not quickly get their desired information from the comments.At the same time,the monotonous commentary sorting form can not satisfy the individual needs of different consumers to comment information.So how to innovate the presentation of online reviews and give high-quality comments to the online consumers first become the hottest research topic.In order to meet the individual needs of different consumers and to select high-quality comments,this paper takes the hotel reviews of Dalian City as an example,research the classifies of reviews.First,the subject of the classification of comments to help consumers quickly locate their own interest in product features;and then under each subject of the sort of comments,make sure that high-quality comments can be given priority to the consumer.Firstly,this paper studies the subject extraction based on the LDA model.In determining the optimal number of topics,this paper uses the PMI to measure the consistency of the subject in this interval,taking into account the fact that the number of subjects is controlled between 5 and 50,so as to determine the optimal number of topics.Considering that a comment may involve more The classification method is compared with the result of manual labeling,and the accuracy,recall rate and F value of different thresholds are compared with each other.The method is based on the following criteria:Determine the optimal threshold.According to the threshold to determine the subject type of the subject,so that the classification of the subject according to the theme.Second,in order to sort the comments on each subject according to the level of effectiveness.In this paper,the above section of the hotel out of the "quality of service"under the theme of the comments,for sorting research.First,the 0/1 mark is valid and invalidated by a certain principle,and the result of the mark is used as the dependent variable to change the validity of the comment into a binary classification problem.Then,the subject probability of the subject-probability matrix in the LDA model is used as the representative feature of the relevance of the subject,and the relationship between the degree of confidence,information richness,comment objectivity,commentary simplicity,reviewer's experience,A total of 28 influencing factors influencing the usefulness of commentary,build a sorting model under the theme of the hotel review.After using feature extraction of random forest,we choose logical regression and support vector machine to construct the method,and compare and verify the method by accuracy.The experimental results show that the accuracy of SVM is higher than that of logic regression in this research.There are two main innovations in this paper:first,the use of sub-theme sort the form,the first online comment by topic classification,then the subject of each comment under the individual validity of the order.Can help consumers quickly locate their own interest in product features.Secondly,in the question of the validity of the subjective review,the subject probability in the subject-probability matrix of the LDA model output is the representative feature of the validity of the comment.It is proved that the feature Which is of great significance to the validity of the comment.