ABSTRACT
Sentiment analysis has become an important research area with the rapid growth of Chinese online reviews. This paper proposes a supervised machine learning method based on Maximum Entropy for the classification of Chinese reviews. Maximum Entropy is a probability distribution estimation algorithm which has been applied to a variety of Natural Language Processing tasks. Our experimental results demonstrate that the classifier based on Maximum Entropy is ideal for the classification of Chinese reviews, which can both raise accuracy and recall rate to considerable levels. Pre-processing of the reviews for the purpose of training and testing is also introduced in this paper.
