Thai Sentiment Resource Using Thai WordNet

Abstract

Language corpus and resources are essential for text analytics research. For popular language such as English, the resources are publicly available. However, Thai resources are hardly found. This work proposes to construct Thai sentiment resource from English-Thai dictionaries, Thai WordNet and SenticNet. Unlike an existing previous work [ 1 ] in which one Thai term is limited to have only one set of sentic values, which is unrealistic. This new methodology proposes that one Thai term can have more than one sentic values based on its synsets. The proposed approach employs synset information from Thai WordNet and hence able to construct many sets of sentic values for Thai terms. The resulting Thai sentiment resource contains 28,780 terms comparing to 16,584 terms from the previous work. Thai terms and their associated synsets can be visually explored via web application.