Dominant emotion recognition in short story using keyword spotting technique and learning-based method

Abstract

Short story is one of the stories which has storyline, idea, thought, and feeling. A story is precisely contained of emotion. That is why emotion can be recognized in a short story. The aim of this research is to recognize the one most dominant emotion in short story. This research used Paul Ekman's basic emotions, which are angry, disgust, fear, happiness, sadness, and surprise. The method that was used in this research was hybrid method, a combination of keyword spotting technique and learning-based method. Keyword spotting technique utilized words in NRC Emotion Lexicon which initially were translated from English to Indonesian. On the other hand, learning-based method used three algorithms, those were Multinomial Logistic Regression, Support Vector Machine (SVM), and Naïve Bayes Multinomial. Those three alghorithms used instance sentences in short story. Hybrid method implemented voting rule to determine the one most dominant emotion. The accuracy of hybrid method resulted in this research was 63.33%.