Comparison of Thai Sentence Sentiment Tagging Methods Using Thai Sentiment Resource

Abstract

Opinion mining aims to determine text sentiment from social network. Thai sentiment resource [ 3 ] can be utilized to tag sentiment of Thai text. Thai sentiment resource consists of Thai terms, for each term a number of English synsets and a set of sentic values and a polarity. However, most terms have many synsets. In order to decide on a sentence emotion, six methods for Thai sentence sentiment tagging were proposed and evaluated. The dataset was obtained from social media comments of an airline service. For the six methods, tagging accuracies ranged from 64.7 to 73.6%. The best method used only adverb and adjective synsets to determine a sentence emotion; while the worst method used only verb synsets.