Emotion Analysis of Japanese Sentences Using an Emotion-word Dictionary

Abstract

In this paper, we first review research on emotion analysis, especially in text using natural language processing technology. Then, we explain our developed emotion-word dictionary and emotion-analysis algorithm that can analyze the emotions of Japanese sentences at a high speed. The dictionary has 6,894 entries, and the meaning of each entry is a 10-dimensional emotion category vector corresponding to 10 types of basic emotions. We also give the evaluation results for the speed and accuracy of analysis by the emotion-analysis system implemented using the emotion-word dictionary and emotion-analysis algorithm. Our emotion-analysis system is more accurate than existing Japanese emotion-analysis systems using dictionaries, and can analyze emotions of Japanese sentences at a high speed (10,000 sentences within 1 second). We further describe how to visualize the emotions extracted by the emotion-analysis system.