Bag-of-words (BOW) is now one of the most popular way to model text in statistical machine learning approaches in sentiment analysis. However, the performance of BOW may remain limited due to some fundamental deficiencies in handling the polarity shift problem. A model dual sentiment analysis (DSA), to address this problem for sentiment classification. First propose a novel data expansion technique by creating a sentiment-reversed review for each training and test review. On the basis of proposed model, a dual training algorithm is designed to make use of original and reversed training reviews in pairs for learning a sentiment classifier, and a dual prediction algorithm to classify the test reviews by considering three sides of one review. We have also made the DSA framework from 2-class polarity (positive-negative) classification to 3-class (positive- negative-neutral) classification, by taking the neutral reviews into consideration. Finally, we develop a corpus-based method to construct a pseudo-antonym dictionary, which removes DSA’s dependency on an external antonym dictionary for review reversion. We conduct a wide range of experiments including two tasks, nine datasets, two antonym dictionaries, three classification algorithms and two types of features. The results demonstrate the effectiveness of DSA in addressing polarity shift in sentiment classification.