Sentiment Categorization Based on Machine Learning for Chinese Microblog
Zhang Qing-qin · Future and Development · 2015
For the proliferated data in Chinese micro blog, the research on sentiment analysis is more useful.How to identify the efficient feature sets for classifier is the key point in the machine learning based methods.For getting the knowledge about different affection with different feature sets, the paper selects 5 categories of features including unigram, syntactic features, Micro blog features, target-based features and lexicon features to be compared and evaluated. The experiments are evaluated on the data of medicine metformin. The results indicate that unigram is the most efficient feature and the best accuracy can be get by combining all the features.