Thai Sentiment Lexicon Construction

Online platforms have become increasingly integrated into daily lives of Thai citizens. Besides being used to query information, online platforms are the medium to exchange opinions and discuss various topics via posts or comments. Since the number of posts and comments can be very large, understanding which topics users are talking about and grasping how users feel about those topics can be difficult and time consuming. Researchers therefore have studied how to automatically analyze posts and comments to extract topics and their sentiments using techniques such as topic modeling and sentiment analysis. There are various approaches to analyze sentiments, and many of them utilize sentiment lexicons, a set of words annotated with sentiment, to help with the analysis. Several researches have constructed sentiment lexicons for English language. This paper aims to construct sentiment lexicons for Thai language using a dictionary-based approach in combination with a crowdsourcing approach. First, words are chosen from a Thai dictionary, mostly adjectives since they usually express sentiment. Next, a web application is developed so that users can help vote whether a particular word is positive, negative, or neutral. After a number of votes have been collected, each word is scored to determine its sentiment. To validate and ensure users are voting truthfully, users are authenticated via Facebook, and simple checks are carried out against answers of each user. This process results in a sentiment lexicons for Thai words and should help form the basis for sentiment analysis of Thai text.