Abstract:

We solved the Chinese review sentiment classification problem via describing and evaluating several machine learning approaches (Nave Bayes, maximum entropy and support vector machines) on some features of the Chinese reviews. The experiment shows the three machine learning methods perform well especially support vector machines, and the accuracy of sentence level is up to 88.26%, the accuracy of review level is up to 91.79%.

Key words: sentiment classification, Nave Bayes, maximum entropy, support vector machines