Abstract

Abstract—Thai sentiment analysis can be analyzed in multiple dimensions: product, price, and shipping. The scores from sentiment analysis can be considered as multi-dimensional trust score of e-vendor which affects the purchase decision of the consumer. A consumer’s review that has different sentiments describing the same dimension in the sentence leads to the incorrect scoring problem in SenseComp – the method to automatically analyze Thai sentiment in the consumer’s review in multiple dimensions with sentiment compensation technique. In this paper, SenseComp using syntactic analysis method (SenseComp2) is
proposed in order to improve the accuracy of the original SenseComp. Thai syntactic structure in the consumer’s review is analyzed before the scores in multiple dimensions are given. The results show that SenseComp2 increases the accuracy of the original SenseComp up to 31.88% in the sentiment analysis of
Thai consumer’s review when there are multiple dimensions and various sentiments in the same sentence.

Index Terms—consumer’s review, multi-dimensional trust, sentiment compensation, syntactic analysis, Thai sentiment analysis