As microblogs have become one of the most important social platforms, it is considered to be extremely valuable to extract user interests hidden behind microblogs. In this paper, we introduce a framework, which is built on the improved TextRank model, to analyze the personal interest of microblog users. In the framework, we first create a catalog of user interests basing on hot tags of Sina Weibo, the largest microblog system in China. And then TFIDF factor is used in TextRank model to deal with pre-processed microblog contents. After ranking and mapping all extracted words into user interests catalog established previously, we get corresponding user interests tags and a user interests model. Experimental results on Sina Weibo data imply that the proposed framework outperforms other existing methods.