Research On Sentiment Analysis Of E-commerce Reviews Based On Latent Dirichlet Model

With the fast expansion of the Internet and information technology,there are generating amount of E-commerce users' reviews,these reviews people's opinion about the object.Text sentiment analysis use the user's reviews as the research object,the task is mining,classification the information of these text sentiment point.This information has important practical value and scientific value in the field of business and science.How to accurately and effectively use the information from the massive reviews is very important for the business.It can provide effective suggestions for the purchase and sale to business.And it can provide all aspects of commodity information to help consumers know more attributes of the products,choose the right products for them.In this paper,we mainly focus on the classification of two kinds of emotions in unlabeled text.If we can calculate the text automatically by computer,analyze what the user comments positive attitude to the product,which users express negative attitude,and the positive degree or negative degree.We can analysis and calculate the text with emotional marks to achieve the classification of unmarked text by computer.This paper proposes an approach called SE-LDA based on Latent Dirichlet Allocation to extract latent emotional sentiment polarity.This approach reduces the dimension of the text base on extraction of the sentiment words and object words.Use the LDA algorithm.calculate the document-topic distribution and the topic-word distribution.Experiments show that the proposed approach has outperforms state-of-the-art techniques in sentiment polarity identification.