Short Text For Multi-classification Of Microblog

Microblog as a new kind of internet application have become a new generation of social media, which spread information efficiently, have many users, content rich information. How to effectively obtain microblogging data and reasonable analysis of the data to discover user attention distribution and emotional tendencies of a focus on the event has practical research value.The main contents are as follows:⑴ Have been put forward for the purpose of the study, to achieve the purpose of the research framework: By combining text classifier, the objective and subjective classifier, emotional tendentiousness classifier for a particular focus on microblog analysis of the dis tribution of emotion.⑵ Have been put forward structured respresentation model of microblog on the basis of analyzing the characteristic of the microblog’s structure to carry out the multi-classification of microblog. In the study of short text classification, Using TFIDF feature selection methods, avoid the large dimension duing to the excessive number of the theme, which can effectively control the characteristic dimension. Combining the features obtained by the TFIDF and theme features obtained by LDA text representation model effectively improve the quality of the LDA text representation model.⑶ In the process of microblog sentiment analysis, first,according microblog observable structure defined microblog processing structure, and then combine multiple existing sentiment analysis- basic dictionary database proposed more comprehensive emotion analysis dictionary database, and through the use of a comparative analysis of the voting mechanism more existing emotional classifier results emotional polarity discrimination classification.⑷ To test the above theoretical results in practical applications, design and implement a complete data obtained from the data processing and storage, data analysis prototype system to the data presented. The system verifies the practical value of the present theory.