Abstract—In this study, a new sentiment analysis model of online-shopping reviews based on hidden Markov model has been proposed. Both the influence of the latest two comments and the most popular comment from the Amazon Japan review page are taken into consideration. The supervised training method is used to train this model, and then the model is optimized by using a variation of genetic algorithm. The performance is evaluated through an experiment of sentiment classification of online-shopping reviews of Amazon Japan’s tea category comparing to other methods from previous ones such as Support Vector Machine, Logistic Regression with built-in cross-validation and so on. The result shows that the adapted hidden Markov model has the highest f1 score among the other baseline methods.
 

Index Terms—HMM, 2 dimensional HMM, machine learning, sentiment analysis, online review

Cite: Xiaoyi Zhao and Yukio Ohsawa, "Sentiment Analysis on the Online Reviews Based on Hidden Markov Model," Vol. 9, No. 2, pp. 33-38, May 2018. doi: 10.12720/jait.9.2.33-38