Content-based emotion classification in online social networks for Chinese Microblogs

Recent years, social networks are popular throughout the whole world. In China in particular, more people spend their time on social networks. Sina Weibo, as the most popular microblogs in China, records millions of microblogs from different population. In this paper, we study and understand sentimental feelings of Weibo by methods of mathematical statistics and analysis. Firstly, we propose a novel three-step extract (NTSE) algorithm to extract meaningful microblogs. Secondly, we identify the similarity of microblogs sent by specific population. Then, we present the naive Bayes algorithm to classify microblogs into three types: positive, negative or objective. For testing the algorithms, we collect Weibo data from specific population of Sina Weibo to form two datasets: student dataset and profession dataset. Some interesting findings include: i) around 20% microblogs are meaningless; ii) only half of microblogs' contents have expressed emotion; iii) students tend to post microblogs with negative emotion among the emotional trends; ix) six professional persons tend to publish positive microblogs. The results of our experiments show that students in five universities in China are more inclined to express negative feelings in the social networks. On contrary, some professional persons including IT, actors and writers and so on more likely to publish positive microblogs.