Abstract
This paper concentrates on the study of opinion classification task. We propose a method to automatically class review with appropriate evaluation category. Our method utilizes extracting features from the training data and uses machine learning algorithm to train an evaluation classification system.We select Maximum Entropy (ME) as our machine learning module. In training time, we utilized the review corpus from the Web and a category dictionary to extract opinion words and phrases as our feature set. And then we used ME to train with feature set to get an evaluation classification module. At run time, a given review is automatically transformed into a feature set and sent to a classification module, and then return a suitable evaluation.