Research And Application Of Microblog Topic Mining Algorithm Based On Sentiment Classification

Nowadays, microblog has been developing rapidly due to the influence of innovative technology, making more and more people comment on the information they focus on through the platform. The platform has become a part of people’s daily life. It produces a lot of information contained in users’ comments, including newsmakers, emergencies, shopping experience and so on. The information not only indicates users’ sentiment orientation, but also contains topic information with important social value and commercial value. Therefore, it is very important to analyze the sentiment orientation of microblog comments and dig out the information that people care about or need, which has become a research focus of scholars at present.This paper focuses on how to dig out the potential, valuable and sentimental topic information from users’ comments on microblog. The main jobs and innovations of the paper are as follows:(1) Given that positive and negative sentiments have totally different effects on topic result, this paper put forward a microblog topic mining algorithm based on sentiment classification(SC-LDA, Sentiment Classification-Latent Dirichlet Allocation). The basic idea is as followings: First, classify the word segmented microblog comments by sentiments use SVM(Support Vector Machine), therefore positive and negative sentiments are obtained. Then use the LDA(latent Dirichlet allocation, LDA) model to dig correlative topics from the classified comments segmentation.(2) Simulation experiment of SC-LDA is done. The result reveals that SC-LDA is not sensitive to the size or sources of the training dataset. The paper also tries to dig topics from the raw data, and it turns out the topic information cannot indicate users’ sentiment orientation and it also affects the inference from the third party, which verity the effectiveness of the proposed algorithm in another way.(3) In order to get a better user experience, this paper designs and implements a microblog topic mining system based on sentiment classification in view of the SC-LDA algorithm. The system can apply to microblog data fetching, data preprocessing, feature extraction, sentiment classification, topic mining and results presentation, etc.