Emoji and Emoticon in Tweet Sentiment Classification

• Published in: 2020 6th Information Technology International Seminar (ITIS) pp. 145 - 150
• Main Authors: Arifiyanti, Amalia Anjani, Wahyuni, Eka Dyar
• Format: Conference Proceeding
• Language: English
• Published: IEEE 14-10-2020
• Subjects: Blogs; Buildings; Classification; Emoji; Emoticon; Faces; Machine learning algorithms; Sentiment Analysis; Social Media; Social networking (online); Support vector machines; Training
• DOI: 10.1109/ITIS50118.2020.9320988

Indonesian Twitter's user tends to use emoji or emoticon when tweeting. Emoji and emoticon can help them for conveying their message efficiently. It is used for expressing their sentiment about a certain topic. For analyzing sentiment with classification technique, the use of emoji and emoticon might take into consideration. In this paper, we knew that Indonesian Twitter's User not only uses emoji in their tweets but also emoticon although emoji used is dominating tweets than emoticon. In some cases, both of them are used simultaneously in a tweet. Emoji and emoticon affect classification model performance positively so that emoji and emoticon might not remove in-text pre-processed and include in classification features. For included in classification feature emoji and emoticon convert into its Unicode Name, but this conversion slightly below classification model from conversion emoji and emoticon into its lexicon sentiment. Conversion emoji and emoticon into its lexicon sentiment make classification model performance more stable. But considering emoticon and emoji cannot simply be interpreted into certain sentiment class so conversion into Unicode Name is encouraging. SVM produces the best performance followed by Bernoulli Naïve Bayes and the last one is Multinomial Naïve Bayes.