A Comprehensive Ranking Method Based On Sentiment Analysis Of Taobao User Evaluation

As one of the largest e-commerce platforms,Taobao has millions of registered users.The daily number of surfing website and concluding a transaction are measured in millions,too.In addition,Taobao's registering requirements for businesses are very low,both large flagship stores,as well as ordinary shops,so that the quality of goods is uneven,and selecting their favorite products quickly on Taobao is very difficult.At the same time,after consumers get the goods,they can give evaluations about the description of goods,service of store and logistics in the evaluation system.The evaluation information reflects the problem consumers caring about.And to other consumers who judge the emotional tendency of the evaluation information can determine whether to purchase the product.Therefore,how to quantify the emotional tendency of evaluation information and how to recommend satisfactory products to consumers are the focus of this article.Aiming at the problems mentioned above,a product recommendation method based on the sentiment tendency of evaluation information is proposed.The whole process involves two aspects.One is to do sentiment analysis on the evaluation information of users,and the other is to calculate the comprehensive scores of other factors of the products.In the sentiment analysis,firstly,select the words in where a strong emotional tendency is from the 1,000 training samples artificially,and use the dictionary named HowNet to build an emotional lexicon;Then,use existing word segmentation tools to separate the evaluation information and mark the words.Extracting feature words from product attributes with apriori algorithm in order to complete the construction of the feature lexicon;Next,through the second tagging of user evaluation information,constructing the phrase model from the training sample in combination with emotional lexicon and feature vocabulary.Calculate the emotional value of each phrase model and form a phrase model library;Finally,each sample is processed to obtain a corresponding emotional value.In the composite score,the value of the product evaluation information is added.And the common factor contribution rate is used to weight the various attributes of the product so that obtain a comprehensive score for each product,which is recommended to consumers in the descending order.In the final order of recommended products,it can be seen that the main focus points are sales number in 30 days,successful transaction volume,evaluation about store,emotional value,which are all related to users' evaluation.We consider the factors such as prices of commodity,sales volume,credit and evaluation instead of Taobao's existing sorting methods.Analyze the relationship between these factors in depth.Then propose a comprehensive ranking method based on users' evaluation,that humanized sorting method can satisfy the shopping needs of customer.