Sentiment Ananlysis On Reviews Of Network Business System

With the development of the Internet,especially the popularity of the intelligent mobile terminal,more and more users through the network platform to purchase and leave reviews.Reviews can reflect the user's preferences and the advantages and disadvantages of goods,a direct impact on the potential customers to purchase,but also for manufacturers and sales companies to provide decision-making.So a variety of sentiment analysis techniques are applied to the field of reviews.How to make the results of sentiment analysis play a role in the actual production is a very worthy of study.The sentences in the reviews are different from those in the forum or microblog.Most of them are short and the subject is ambiguous.Therefore,the study of reviews based on semantic rules can achieve better results,but the existing semantic rules can not satisfy the specific structure in the reviews,which needs improvement.First,the paper introduces the theory,background,current situation and related technology of sentiment analysis.Then,based on the review of electronic commerce in the field of apparel,a sentiment analysis process based on semantic rules is proposed.The paper summarizes the rules of Chinese sentence structure and a method to calculate the intensity of sentiment tendencies is proposed.And then a sales forecasting model to study the relationship between the intensity of sentiment tendencies and sales is proposed.Classification algorithm is used to analyze the optimal number of pages for the reviews that affect sales.Finally,a case study is used to validate the research results.The paper uses the language of Python.First,the original corpus is constructed by crawling the 4000 items of reviews data from two different goods on Taobao,and the original corpus is segmented and annotated by using the Language Technology Platform of Harbin Institute of Technology to generate XML corpus.Secondly,using the XML corpus and general sentiment dictionary to establish sentiment dictionary in the field of apparel and use the dictionary to modify the XML corpus.And then use the proposed semantic computing rules to calculate the intensity value of sentiment tendencies of each review.Furthermore,Bayesian and other seven classification algorithm are used to analyze the relationship between the intensity of sentiment tendencies,the number of pages of reviews and sales.And recall rate and precision are used to evaluate and determine the the optimal number of pages for the reviews.Finally,the relationship between the intensity of sentiment tendencies and the sales volume was analyzed by using the linear regression model,the neural network model and the support vector machine regression model according to the optimal number of pages for the reviews.