Detecting Emotions of Indonesian Songs Based on Plutchik’s Theory using Data Mining

Abstract

Listening to songs is a daily activity that everyone engages in. Most people choose songs based on their mood, so a system is needed to detect emotions from song lyrics. Previous research only focused on five basic emotions: happy, sad, love, anger, and fear. In this study, we propose a new method to detect emotions from song lyrics using Plutchik's emotion theory. The data used for this research consisted of 250 song lyrics from Indonesian songs. This research categorizes human emotions into eight: joy, trust, surprise, sadness, disgust, anger, and anticipation. Next, the threshold value is calculated. This value is used to determine the dominant emotion. If the frequency value of an emotion is higher than the threshold value, the system considers it as the dominant emotion. The dominant emotions are then classified into positive and negative emotions using cosine similarity calculations. The sampling technique involves using 30% of the test data, resulting in an accuracy of 0.81.