Traditionally, text sentiment analysis is only limited to sentiment classification. The review is simply divided into two types: positive and negative comments. In practice, we sometimes need to rank the review, which cannot be solved by traditional sentiment classification methods. To solve this problem, this paper proposes a novel review text sentiment ordinal regression based on Latent Semantic Index. Firstly latent semantic indexing is used to extract features for review texts and then an ordinal regression method is used for review text sentiment analysis. The experimental results on Movie Reviews database proved the effectiveness of the proposed method.