The media feature analysis of microblog topics

Abstract

As microblogging grows in popularity, many research articles are exploring and studying the micro-blogs, especially the English micro-blogging, i.e., twitter. However, Chinese micro-blog service starts rather late and a few research is about its data and characteristics. In this paper, we give out the media feature analysis of microblog topics. Firstly we present our observations tweets and users from Sina by crawling 14 topics and their 74,662 tweets and give out the topic evolution in a certain interval. Then considering the microblogs under a topic exist a lot of redundant information, so in order to reduce the trainning datasets for studying on microblogging , we respectively select diffierent data source as our datasets and give out the evaluation method. We have also studied the microblog semantic extraction based on the topic model of LDA (Latent Dirichlet Allocation). we conclude that the active period of most micro-blog topics is about a month and the out-dated topics will be replaced by the upcoming and related new topics and find that the tweets that appeared in the peak time or the tweets from authenticated users can reflect the whole tweets situation of a topic. However, due to the microblog text is so short that LDA for semantic extraction is not ideal. © Springer-Verlag 2013.