The sentiment analysis of the review texts of micro-blog is helpful for deep mining Chinese Micro-blog (Weibo)-one of the main social media in China. Aiming at the shortcomings of the widely used machine language in sentiment analysis of texts when dealing with sentences containing connectives, this paper formulates rules for dealing with Chinese connectives, incorporates expression symbols into feature vectors, calculates sentiment decision scores with sentiment dictionaries, and proposes an enhanced supervised learning model that is based on language rules and emotional scores. Examples show that the proposed model can significantly improve the effectiveness of text classification.