COMPARISON BETWEEN THE HUMAN EMOTION TRANSFER RATIO AND THE SIMILARITIES OF EMOTION
Existing methods to recognize emotions of a sentence can recognize a few kinds of emotion, and many methods use a dictionary that emotion weights is related every words. Our aims are to propose a method which can recognize with high precision using feeling expression and can add new recognizable emotion easily. To realize the proposed method, we proposed an emotion similarity calculation. This method calculates similarity between emotions that two sentences represented. In the emotion similarity calculation, emotion similarity between an input sentence and classi.ed sentence by emotions is used. The classi.ed sentences are in emotion corpora. The calculation formula is based on BLEU which is a machine translation evaluation method. In this paper, an emotion similarity calculation method using N-gram frequency dictionaries is shown. This method can recognize with high precision compared to the past one. The N-gram frequency dictionaries are made from each emotion corpora. To examine the recognition accuracy with our method, we compared the precision and the human emotion transfer ratio between humans. As a result, we found the precision is 89.91％ of the ratio. This result means the recognition with our method can recognize like a human.
affective computing similarity calculation corpus-based
Kenichi MISHINA Seiji TSUCHIYA Fuji REN
Graduate School of Advanced Technology and Science,The University of Tokushima Minami josanjima,Toku Institute of Technology and Science,The University of Tokushima Institute of Technology and Science,The University of Tokushima;School of Information Engineering,Be
国际会议
2008高等智能国际会议(2008 International Conference on Advanced Intelligence)
北京
英文
2008-10-18（万方平台首次上网日期，不代表论文的发表时间）