Sentiment Feature Identification from Chinese Online Reviews

Using statistical machine learning methods to select features and reduce dimensionality for sentiment classification of Chinese online reviews based on hotels. The results showed that: sentiment classification of Chinese online reviews will obtain the highest accuracy when taking adjectives, adverbs and verbs together as the feature. The best reduction method is Document Frequency, followed by Chi-square Statistic, and Information Gain is the worst. Mutual Information is not suitable for the sentiment classification of Chinese online reviews while Factor Analysis can be a potential method though merely used.