The Research Of Sentiment Multi-Classification Based On UGC

With the development of the Internet and the popularity of Web 2.0 applications, UGC(User Generated Content) presents an explosive trend.Sentiment Analysis of the UGC gradually reflects the enormous commercial and academic value. From the current research situation, the sentiment multi-classification of the text, especially for UGC sentiment multi-classification problem is still a difficult research. Based on this, this paper focuses on the sentiment multi-classification problem of UGC.This paper analyzes the characteristics of UGC text data, then designs and implements a system of data preprocessing which can clean and process the original UGC text to meet the needs of other analysis tasks,and the analysis of UGC text data can also be migrated to other areas. For the UGC single sentiment multi-classification problem, this paper improves the structure of the traditional network model by combining the advanced learning technology of the current frontier. Based on the previous work, this paper proposes an L-RNN model, which can make up for the shortcomings of other network models while retaining other depth learning models and more conducive to retain and use of the original text of the information. The stability and effectiveness of the L-RNN model are verified by the experimental results. Considering the complexity of the UGC emotion in the real world, this paper studies the complex emotion multi-classification of UGC by combining the cognitive emotion model in the field of psychology and proposes a Simple-OCC model suitable for UGC emotion partitioning. An improved multi-label learning model, HML-kNN, is also proposed. Through the combination of Simple-OCC and HML-kNN model, the UGC complex sentiment multi-classification task has achieved good results.