Building thesaurus lexicon using dictionary-based approach for sentiment classification

Abstract

Sentiment classification categorizes people's opinions from the data. Nowadays, people express their personal interests, feelings, and opinions on social media, and the posts on social media are frequently used as the data for sentiment classification. One of the sentiment classification approaches is a dictionary-based approach. A traditional dictionary-based sentiment classification approach uses word matching based on the lexicon. However, many posts cannot be analyzed by traditional dictionary-based sentiment classifier due to the absence of the sentiment words in the lexicon. For this reason, it is needed to expand the lexicon so that the lexicon can contain the words. In this paper, we propose a method to build thesaurus lexicon using dictionary-based approach for the sentiment classification. The proposed method uses three online dictionaries to collect thesauruses based on the seed words, and stores only co-occurrence words into the thesaurus lexicon in order to improve the reliability of the thesaurus lexicon. Also, this method recursively collects thesauruses which are a set of synonyms and antonyms to expand the thesaurus lexicon. This recursive thesaurus collection provides effective expansion of the lexicon from small set without the use of human resource, and the expanded thesaurus lexicon is used to increase availability of posts and used to increase accuracy of the sentiment classification.