Sentiment classification is an automatic opinion classification method to classify the product reviews on web into positive or negative opinions to help consumers or sellers to understand the opinions and evaluations from existing customers. Semantic-oriented approach is one of the recent developments in sentiment classification. Up to now, most research of sentiment classification is on English reviews,and little work has been done on Chinese reviews using sentiment classification. The detailed techniques used in English review cannot be applied directly to Chinese reviews due to the different characteristics between these two languages. This study modified and improved the semantic-oriented approach to a 6-step process for Chinese review, focusing on the modification and improvement on the text segmentation and reference words pairs (RWPs) identification. Two experiments were conducted on book reviews and cell phone reviews. The results show that the performances of the proposed approach are comparable to those of the existing English reviews classification studies.