Opinion Mining Research Of LDA Maximum Entropy Model Based On Cloud Model Theory

With the fast development of the Internet, our society entered the era of "Internet+" and "big data". The unprecedented prosperity of the e-commerce and online social media make people have more chances to show their opinions on the Internet, online comments contains a lot of valuable information, these information can serve businesses, customers and government departments. But confront the massive online review data of user’s network behavior provided by Internet companies, relying on an artificial way can not effectively extract the key information. In this environment, opinion mining technique provides a new way to solve these problems, fine-grained opinion mining can provide more detailed information for users, so it get extensive attention of the domestic and foreign researchers.Opinion mining has three main tasks:sentiment classification, opinion extraction and opinion analysis, in order to complete the above tasks, the researchers proposed a lot of methods, these methods can be simply summarized as rule-based method and statistical machine learning method. Early opinion mining researches are mostly based on rule-based method. Experts is needed to define characteristics of words and rules in different areas in rule-based method, although in some extent it can suits opinion mining’s needs, but the characteristics of heavy workload and poor adaptability make this method is not suitable for massive data mining. In the perspective of statistical machine learning based on statistical machine learning, which is represented by the topic model, has been recognized by researchers for its domain adaptability and non supervisory.LDA Model is a kind of unsupervised learning topic model, which expresses a document as a collection of words. It resembles a bag of words that contains mutually independent unrelated words. Then we utilize techniques like Naive Bayesian methods, Expectation Maximization Algorithm and Gibbs Sampling to approximately reason document-topic distribution and topic-word distribution and thereupon we then obtain related perspectives. The position of words and semantic relationships have profound effect on the expression of perspectives in the particular document. To acquire more specific and valuable information about the perspectives, improving the traditional LDA Model before applied to fine-grained opinion mining is essential.Furthermore, natural language is frequently ambiguous and correlated resulting in perspective mining and modeling uncertainty. On the one hand, the line of demarcation while describing qualitative concept is fuzzy and on the other hand, the quantitative expression of models is of randomness. The common used LDA Model that applied to perspective mining only considers the randomness of quantitative expression while ignores the fuzzy line of qualitative concept demarcation. In order to overcome these shortcomings, this article use Maximum Entropy Model and Cloud Model and Theories to improve the standard LDA Model aiming atfine-grained opinion mining of online remarks. First of all, the introduction of Maximum Entropy Models makes full use of the position of words and semantic information so that more detailed words’division of the document would be made. Secondly, after applying Cloud Model and Theories to Maximum Entropy LDA Model, sentiment correction algorithm can be achieved ascribing that Cloud Model could accomplish qualitative and quantitative conversion towards uncertain sentiment in the document. The similarity computation of global emotion and subject emotion would be achieved through the lines of approach degree and cloud expectation and sentiment deviation would in turn be revised, moreover accuracy of perspective mining would be Effectively improved.Simulation experiment was conducted and visualized results are attached to the end of this article. The experimental results confirmed the effectiveness of the theory proposed.The research background and research significance, current research status have been introduced in the first chapter. The second chapter introduces the theoretical basis of opinion mining, and lists the basic mathematical knowledge and models which is involved in this paper. In the third chapter, Sentiment Cloud Maximum Entropy LDA Model is introduced in detail, and the related modeling ideas, theoretical deduction and sentiment revision algorithm are given. The fourth chapter mainly introduces the process of simulation experiment and the analysis of related results, and the experimental results are visualized, which verifies the validity of the theory. In the fifth chapter, the related research topics are summarized and prospected.