Using Prefix Tree To Improve The Performance Of Chinese Sentiment Analysis

As social media services (e.g. Wikipedia, Facebook, Twitter, Linkedin, and so on) become more and more popular, it is of greater research interest to raise the efficiency of using Sentiment Analysis to predict future opinion trends. Based on the naive Bayes classifier, this research proposes a novel emotion classifier, CCLM (Combined CKIP Language Model), to enhance the precision of opinion classification. Due to the fact that CCLM tends to produce a larger size of item sets than other language models, the prefix tree is adopted to improve the overall performance. This study adopts short messages retrieved from Plurk to explore sentiment analysis of Chinese texts, while making additional effort to observe change in execution time after adding the Prefix Tree.