Designing Korean Emotion Analysis System Combining KoBERT Model and Ensemble Machine Learning Model Considering Emotion Words

Emotions can be expressed through language, facial expressions, voice, and other means. However, facial expressions and voice can be hidden or distorted by an individual's will. However, language has less distortion of emotions due to the situation and linguistic meaning. Therefore, in this paper, we propose a Korean emotion analysis system using only linguistic expressions. The proposed system considers the emotional elements of words. The emotion word dictionary considered the characteristics of the Korean language by using FastText, which learns subwords using n-grams. The proposed system proposes a deep learning module learned through KoBERT Fine-tuning and a sentiment analysis module through machine learning ensemble machine learning for sentiment analysis. This paper improves the accuracy of the Korean emotion analysis system by combining deep learning modules and machine learning modules.